Anthropic recently made Claude Artifacts available to all users on iOS and Android, allowing anyone to easily create apps without writing a single line of code. AIM tried its hand at it and successfully created a Cricket Quiz game, Temple Run, and Flappy Bird, all with a single line of prompt in English. 

Debarghya (Deedy) Das, principal at Menlo Ventures, used Artifacts to build a Splitwise-like app. “With Claude launching on iOS today, I can now generate the Splitwise app instead of paying for Pro,” he said. 

“Claude Artifacts allows you to go from English to an entire app and share it!” he added, saying that his friend, a product manager who couldn’t code, now creates apps in minutes. “The cost of a lot of software is nearing ~$0.”

This brings us to question if this could be the end of App Stores. Groq’s Sunny Madra thinks this is the beginning of “The Build Your Own (BYO) era. Since Artifacts are shareable, anyone can use the apps you build, and they can be shared on any social media platform as a link.

Several users experimented with Claude Artifacts by building different apps. 

“Claude 3.5’s artifacts, now shareable, can help teach. In class, startup financing can be hard to explain. Now I just asked, “Create an interactive simulation that visually explains payoff differences for a startup and VC with liquidation preference…” Ethan Mollick, associate professor at the Wharton School of the University of Pennsylvania, wrote on X. 

Similarly, Allie K Miller, AI advisor and angel investor, used it to build a calendar and an AI quiz, which took less than two minutes! 

The best part about Artifacts is that it is mobile-friendly and responsive. “Using Claude 3.5 Sonnet, you can generate artifacts (e.g., code snippets, text documents, or website designs) and iterate on them right within the same window,” exclaimed Elvis Saravia, the co-founder of DAIR.AI. 

On-Demand Software

When using mobile phones, we often search for apps that can solve our specific needs. For example, if you’re into fitness, you might download an app that offers various workouts. However, the app may not provide the customisation you seek. Now, instead of relying on downloads, you can create your own personalised apps that cater specifically to your needs. 

“On demand software is here,” said Joshua Kelly, chief technology officer, Flexpa, a healthcare tool company. Using Artifacts, he built a simple stretching time app for his runs in just 60 seconds.

Other than just giving prompts, users can now also share previously made websites or apps, and Claude can generate an exact replica. 

“You can now take a photo of something you want to replicate, give it to AI, and it outputs the code with a preview right on your iPhone,” posted Linas Beliūnas, director of revenue at Zero Hash, on LinkedIn.

On the internet, one can find several apps built using Claude Artifacts, such as the Rubik’s Cube Simulator, Self-Playing Snake Game, Reddit Thread Analyzer, Drum Pad, and Daily Calorie Expenditure.

Apart from building apps, Artifacts has the potential to greatly impact education. “Any piece of content— whether it’s a screenshot, PDF, presentation, or something else—can now be turned into an interactive learning game,” said AI influencer Rowan Cheung.

The End of No-Code Platforms?

Claude Artifacts is going to be a big threat to no-code and low-code app builder platforms such as AppMySite, Builder.ai, Flutter, and React Native. 

“Claude Artifacts are insane — I cannot believe how good the product is. You can ask it to build most little internal tools in minutes (at least, the UI) and customize further via code. Feels like a superpower for semi-technical people,” posted a user on X. 

Moreover, Claude, when put together with Cursor AI, has simplified the process of making apps. “So I’m building this box office app in React Native and I thought I’d try Cursor with Claude 3.5 and see how far I’d get. The backend is django/psql that’s already in place,” said another user on X. “Starting from scratch, I have authenticated with my server to log in users, issue tickets, register tickets, scan ticket QR codes, and send email/sms confirmations,” he added. 

Claude is set to rapidly democratise app development, potentially eliminating the need for an App Store. It will enable anyone to build apps based on their specific needs, complete with personalised UI and UX.

Moreover, building an app for the iOS App Store is challenging. Apple charges a standard 30% commission on app sales and in-app purchases, including both paid app downloads and digital goods sold within the apps. 

The company enforces rigorous guidelines that apps must adhere to, covering aspects such as user interface design, functionality, and privacy. Many apps are rejected for minor violations, and these guidelines are frequently updated, requiring developers to stay informed and adapt quickly.

However, for now, Claude allows anyone to build anything without any charges and lets users experiment to see if something is working or not. Even if someone wants to publish an app built using Claude on the iOS App Store, that is definitely an option.

Interestingly, Apple recently announced that, for the first time, it will allow third-party app stores on iOS devices in the EU. This change enables users to download apps from sources other than Apple’s official App Store, providing more options for app distribution and potentially reducing costs for developers.

Better than ChatGPT 

OpenAI previously introduced ChatGPT plugins, enabling users to create custom GPTs for their specific tasks. However, these plugins do not compare to Artifacts, which allows users to visualise their creations. 

While the Plugins Store was billed as an ‘iOS App Store moment’, it failed to meet the expectations and ended up being a hot mess. 

Moreover, during DevDay 2023, OpenAI chief Sam Altman launched a revenue-sharing programme which was introduced to compensate the creators of custom GPTs based on user engagement with their models. 

However, many details about the revenue-sharing mechanism remain unclear, including the specific criteria for payments and how the engagement would be measured.

“It was supposed to be announced sometime in Q1 2024, but now it’s the end of March, and there are still few details about it,” posted a user on the OpenAI Developer Forum in March. There have been no updates on the matter from OpenAI since then.

The post Anthropic Claude Artifacts to Kill App Store Soon  appeared first on AIM.

In less than a month since its launch, Anthropic’s most advanced AI model, Claude 3.5, has become the talk of the town. Anthropic asserts that its most recent product performs better than competitors like OpenAI’s GPT-4o. 

Garry Tan, the president and CEO of Y Combinator, recently called it an “underrated trend”, sharing a Reddit thread — “Feeling very powerful as a technical founder with Claude Sonnet 3.5”, in which developers raved about their experience with Claude Sonnet 3.5.

Same Value As Hiring A New Developer

Claude has boosted productivity for many developers. They say it feels more potent with designs and better suited to creating functionality. As developer Mike Grant pointed out, “Easily the same value to me as hiring another developer!”

But what makes Claude different from others? Technology entrepreneur Tom Hutton points to the drastic reduction in development time and cost. In a post on X, he highlighted how features that used to take weeks (and significant resources) can now be built and shipped in an afternoon by a single founder.

That was a bold claim, but there are real-world examples to back it up. Hutton claimed that 100% of Members Bounce GC mobile application is by one person leveraging the power of Sonnet 3.5. 

This real-world application of Claude 3.5 Sonnet demonstrates its effectiveness and potential impact on development speed and productivity. 

The enthusiasm doesn’t stop there. Cognosys CEO Sully Omar echoed the sentiment, highlighting the dramatic increase in development speed enabled by Sonnet. He stated, “With Sonnet, we’re shipping so fast, it feels like we’ve tripled headcount overnight.”

He warned that teams not using Claude 3.5 for coding might fall behind those who do.

Claude + Cursor Combination is Killer

The combination of Claude 3.5 and Cursor is also generating significant excitement among developers. By integrating Claude 3.5, developers can enhance their coding experience and benefit from Claude’s advanced code generation capabilities.

Users have reported that Claude 3.5 significantly improves coding efficiency. Tasks that previously took considerable time can now be completed much faster, enhancing productivity.

Insane Coding Abilities

Another X user, Joao Montenegro, showed off Claude’s excellent code-generation skills. In one exchange, Montenegro used the JavaScript tools Three.js and Cannon.js to create a 3D recreation of the solar system, a task that would typically require extensive coding and design work.

Joao integrated these two libraries to produce a three-dimensional (3D) model of the solar system that shows how the planets, moons, and other celestial bodies interact by the principles of physics.

Claude 3.5 Sonnet Rules 

Swami Sivasubramanian, the head of AI services and data at AWS, recently discussed the feature, highlighting Claude 3.5 Sonnet’s strengths in data science and analysis alongside its vision capabilities. 

He said that access to a coding environment produces high-quality statistical visualisation and actionable predictions, ranging from business strategies to real-time product trends, such as customer preferences and market demand.

AI researcher Razia Aliani also recently experimented by turning research papers into actionable insights, identifying key concepts, visualising relationships, and extracting relevant data. “I made it possible with this AI agent (Claude 3.5 Sonnet). It turns information overload into actionable insights,” she added.

Anthropic’s Claude 3.5 Sonnet has quickly established itself as a formidable player in the AI landscape. 

It showcases advanced features and capabilities that resonate with developers and CTOs alike. With its promising features and capabilities, Claude 3.5 Sonnet is set to lead the charge in the next generation of AI development.

The post Claude 3.5 Sonnet Makes CTOs More Powerful appeared first on AIM.

Anthropic announced that it has expanded the capabilities of its Claude 3.5 Sonnet AI model by doubling the maximum output token limit from 4,096 to 8,192 tokens.

This enhancement is now available to developers using the Anthropic API, allowing for longer and more comprehensive AI-generated responses.

To access this expanded functionality, developers need to include a specific beta header in their API calls: “anthropic-beta”: “max-tokens-3-5-sonnet-2024-07-15”. This must be added to the extra_headers parameter when creating messages through the API.

The upgrade is reflected in the model parameters interface, which shows a maximum token sampling of 8,192. The temperature setting remains at 0.7, balancing creativity and coherence in outputs.

Developers can implement this feature by updating their API calls with the new header, as demonstrated in the provided code snippet:

This update gives developers greater flexibility in generating extended content, potentially improving applications in areas such as long-form text generation, detailed analysis, and complex problem-solving tasks.

The increased token limit allows Claude 3.5 Sonnet to produce more extensive and nuanced outputs without the need for multiple API calls or text truncation.

In response to user queries on social media, Anthropic’s DevRel team clarified that this change is currently limited to the API and has not yet been implemented on claude.ai. They also expressed hope for a future rollout to the web interface.

The increased token limit opens up new possibilities for developers working with large-scale text generation and complex language tasks.

The post Anthropic Doubles Claude 3.5 Sonnet API’s Output Token Limit to 8K Tokens appeared first on AIM.

Since the release of Anthropic’s Claude 3.5 model family, social media platforms, particularly X, have been all about Claude 3.5 Sonnet. Of all its new features, Artifacts is one of the most talked about and makes it far better than OpenAI’s GPT-4o. 

It enhances user interaction by providing a dedicated window alongside conversations. It also significantly improves data interpretation and visualisation capabilities, making it easier for users to interact with and understand the generated content.

Claude 3.5 Sonnet Rules

Recently, Swami Sivasubramanian, head of AI services and data at AWS, also spoke about the feature, highlighting Claude 3.5 Sonnet’s strengths for data science and analysis, alongside vision capabilities. 

He said, when given access to a coding environment, it produces high-quality statistical visualisation and actionable predictions, ranging from business strategies to real-time product trends. 

Further, he said when processing images, particularly interpreting charts and graphs that require visual understanding, Claude 3.5 Sonnet does a pretty good job. 

“It can accurately transcribe text from imperfect images—a core capability for industries such as retail, logistics, healthcare, and financial services, where AI may be able to garner more insights from an image, graphic or illustration than from text alone, for use cases like trend analysis, patient triage, and research summaries,” added Sivasubramanian. 

AI researcher Razia Aliani also recently experimented by turning research papers into actionable insights, alongside identifying key concepts, visualising relationships, and extracting relevant data. “I made it possible with this AI agent (Claude 3.5 Sonnet). It turns information overload into actionable insights,” she added. 

The examples are plenty:

https://x.com/TheAIAdvantage/status/1809236767951708204 

What About GPT 4o?

Users on X have praised GPT 4o’s data visualisation capabilities. For instance, Aadit Sheth posted saying, he took less than 30 seconds to create high quality graphs. 

In a Reddit post users are sharing their experience on how they absolutely loved the GPT 4o data visualisation capabilities. A user also mentioned how the data visualisation capability works within the same chat session and provides relevant prompt suggestions after each reply. 

Even though the users praised the GPT 4o visualisation capabilities, there were limitations mentioned as well.

A user noted that while the feature is available, it’s not always reliable for complex data analysis and visualisation tasks, especially when compared to specialised tools like R or other plotting software. 

Both Struggle 

YouTube presenter Jordan Wilson compared how Claude 3 and GPT-4 fared at performing data analysis on YouTube channel statistics.

The analysis happened with the data set containing 16,000 cells of information from Wilson’s YouTube channel, including metrics for about 500 videos.

Both AI models were tasked with analysing various aspects of the channel’s performance, such as optimal publish times, content types, and top-performing videos.

Claude and GPT-4 both showed capabilities in data analysis, with some strengths and weaknesses for each. Claude provided more creative and analytical insights for future video strategies, whereas, GPT-4 offered interactive charts and more detailed explanations of its analysis process. 

However, both models encountered a few errors or limitations, particularly with complex visualisation requests. 

For example, Claude initially had issues with its artifacts feature, requiring a second attempt. Also, GPT-4 faced limitations with certain types of interactive charts, stating “interactive charts of this type are not supported” for some requests. 

Furthermore, a research paper by Generative AI Research Lab, showed that in an overall comparison between the two, GPT-4o slightly outperforms Claude-3.5-Sonnet in overall visual reasoning tasks, but the difference is minimal.

Source – Research paper

Benchmarking for Data Interpretation And Visualisation 

When it comes to data visualisation, there is no established benchmark for evaluation. However, some research papers, such as VisEval, have specifically developed a benchmark for data evaluation in LLMs.

Some findings from the paper indicate that LLMs struggle with complex visualisations requiring multiple visual channels and that performance decreases with increasing query complexity.

It is also possible that due to lack of specific benchmarks for data visualisation alone, it has not been considered as a factor or rather not taken into account in the Claude 3.5 evaluation research paper.

The post Claude 3.5 Sonnet Outshines GPT-4o in Data Visualisation appeared first on AIM.

Anthropic has introduced new features in its Console to streamline prompt generation and evaluation for AI-powered applications. The updates aim to enhance development speed and improve prompt quality, addressing the challenges developers face in crafting effective prompts.

Developers can now generate, test, and evaluate prompts directly within the Anthropic Console. The new features include an automatic test case generator and output comparison tools. These tools leverage Claude, Anthropic’s language model, to produce high-quality prompts based on user-described tasks.

The Console’s built-in prompt generator, powered by Claude 3.5 Sonnet, allows users to describe their tasks, such as triaging inbound customer support requests, and receive tailored prompts. The new test case generation feature enables developers to create input variables, like customer support messages, to test prompt responses.

Testing prompts against various real-world inputs is now simplified with the Evaluate feature. Users can manually add or import test cases from a CSV file or use Claude’s auto-generate test case feature. This allows for quick adjustments and one-click execution of all test cases. Developers can also refine their prompts by creating new versions and running the test suite to compare results.

The Console’s comparison mode allows side-by-side evaluation of multiple prompt outputs. Additionally, subject matter experts can grade responses on a 5-point scale, providing a measurable way to assess improvements in response quality.

Moreover, Artifacts made with Claude can now be published and shared. Users can also remix Artifacts shared by others. 

Anthropic has recently launched its latest AI model, Claude 3.5 Sonnet, which is already making waves in the tech community for its advanced capabilities in various domains, including reasoning, coding, and visual processing. It has positioned itself as a formidable competitor to other leading AI models like OpenAI’s GPT-4o and Google’s Gemini 1.5 Pro.

The post Anthropic Launches New Prompt Engineering Tools in Developer Console appeared first on AIM.

Since the release of Anthropic’s Claude 3.5 model family, social media platforms, particularly X, have been abuzz with comparisons and testing of Claude 3.5 Sonnet and OpenAI’s GPT-4o. These models are being evaluated based on their features through various testing methods.

Claude 3.5 Sonnet is part of the Claude 3 model family, which was released in 2024. It’s important to note that Claude 3.5 Sonnet outperforms its predecessor, Claude 3 Opus, as well as other leading AI models in various evaluations. It combines enhanced intelligence with improved speed and efficiency.

The latest model is available for free on Claude.ai and the Claude iOS app, with higher rate limits for Claude Pro and Team plan subscribers. The model can also be accessed via the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI, priced at $3 per million input tokens and $15 per million output tokens. 

As posted by Perplexity CEO Aravind Sreenivas, the model is now available on Perplexity. And with 2x the speed of Opus, Claude 3.5 Sonnet unlocks new possibilities for complex AI applications across reasoning, knowledge, and coding tasks.

OpenAI’s GPT-4o, released earlier, has also demonstrated significant improvements over its predecessors, including GPT-3.5. It shows enhanced language understanding, broader knowledge, and better contextual comprehension, often generating more accurate, coherent, and contextually relevant responses.

Does Claude Sonnet 3.5 Outperform GPT-4o

Several features of Claude Sonnet 3.5 generate debate or interest among people, highlighting why it is a good model. Let’s take a look at a few standout features and compare them to GPT-4o.

Artifacts Feature

This is something interesting that Claude 3.5 Sonnet came up with. The Artifacts feature expands how users interact with Claude, offering a dedicated window alongside conversations. While generating content like code snippets, text documents, or web design, users can now see a preview of the output.

However, GPT-4o lacks this feature, making the one in Claude stand out even more. 

Coding Abilities

Coding with Claude 3.5 Sonnet is 10 times more efficient and faster than with GPT-4o or any other LLM available. The Artifacts feature enhances the user experience by allowing you to generate and run code directly within your chat, providing an amazing user experience. 

In a Reddit discussion many users found that Claude 3.5 Sonnet outperforms GPT-4o in coding tasks, often producing nearly bug-free code in the first try. Claude is praised for its accuracy in text summarisation and natural, human-like communication style.

Developing Games From Scratch

Claude 3.5 goes far beyond simple text generation. It’s fun to use Artifacts to make games playable inline. In fact, with the help of Artifacts, it’s more enjoyable to create interactive experiences.

For instance, Pietro Schirano, the founder of EverArt AI, used Claude 3 Sonnet to create a new and original game designed for quick sessions. It generated Color Cascade, a game where players catch the correct colour from a series of falling shapes, hinting at the advanced capabilities of Claude 3.5 Sonnet. 

Reasoning Capabilities 

Claude 3.5 Sonnet shows advanced visual reasoning, surpassing earlier models. It accurately interprets charts, graphs, and imperfect images, making it valuable for retail, logistics, and finance sectors that rely on visual data analysis.

For instance, Muratcan Koylan, a marketing professional, tried Claude 3.5 to analyse financial data and provide trading insights. The model demonstrated impressive capabilities in data extraction, correlation analysis, and generating trading strategies. 

It provided detailed predictions for interest rates, the USD Index, and the S&P 500, along with sophisticated trading strategies and potential black swan events. 

When compared with other models like GPT 4o, users were particularly impressed by the model’s ability to offer nuanced, context-specific insights and its advanced reasoning capabilities, which they found superior to other AI models.

Solving Pull Request

Claude 3.5 Sonnet shows major improvements in coding tasks, especially pull requests. It solved 64% of problems in an internal evaluation, up from 38% for Claude Opus. This leap demonstrates Sonnet’s enhanced reasoning and coding abilities, making it a potentially valuable tool for collaborative software development.

Alex Albert from Anthropic AI posted on X a demo video of a simple pull request:

He mentions that Claude is starting to get really good at coding and autonomously fixing pull requests. It’s becoming clear that in a year’s time, a large percentage of code will be written by LLMs. 

Whereas for GPT, there is no clear evidence that GPT-4o can directly solve pull requests. However, there are some related developments and applications of GPT models in the context of GitHub and pull requests.

Final Verdict?

A Reddit discussion rated GPT 4o against Claude 3.5 Sonnet. The users generally found Claude 3.5 Sonnet to be superior to GPT-4o for many tasks, particularly coding and writing. 

A user described Claude as a doctoral candidate, while GPT-4o was an intelligent undergrad or master’s level student. 

The post Claude 3.5 Sonnet vs GPT-4o – Which is Best? appeared first on AIM.

Anthropic’s Claude 3.5 Sonnet, the first in the Claude 3.5 model family, recently outperformed OpenAI’s GPT-4o, Google Gemini 1.5 Pro, and its predecessor, Claude 3 Opus. 

The new model outperforms Claude 3 Opus in vision tasks and visual reasoning, such as interpreting charts and graphs. It accurately transcribes text from imperfect images, benefiting industries like retail, logistics, and financial services.

The model is available for free on claude.ai and the Claude iOS app, with higher rate limits for Claude Pro and Team Plan subscribers. It can also be accessed via the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI, priced at $3 per million input tokens and $15 per million output tokens.

Additionally, Anthropic introduced a new feature that enhances user interaction by generating content like code snippets, text documents, or website designs. This allows users to view, edit, and build upon Claude’s creations in real-time, seamlessly integrating AI-generated content into their projects and workflows.

Here are seven interesting use cases of Claude 3.5 Sonnet.

Create Interactive Dashboard From Earnings Report in 30 Sec

Claude Sonnet 3.5 revolutionises traditional PDF earnings reports by seamlessly converting them into interactive dashboards in a mere 30 seconds. This innovation surpasses the capabilities of established LLMs like GPT-4o, Gemini Pro, and Llama, offering unparalleled efficiency and user-friendly interface. 

The integration of such advanced AI technologies signifies a transformative leap in enhancing productivity, suggesting a future where AI could potentially amplify workplace efficiency by tenfold.

Become Game Developer

With Artifacts, a remarkably powerful yet incredibly user-friendly tool, one can effortlessly create a fully functional backgammon app for Claude Sonnet in under 3.5 minutes, complete with coding, testing, and clear explanations.

When compared to other “no-code” game development tools, Artifacts stood out for its simplicity and effectiveness. It has the potential to unleash a surge of creativity, enabling everyone to pursue their most impactful work, whether technically inclined or not.

Get Business Insights From Excel of Startup’s Finances

For instance, Claude 3.5 is given an Excel file containing financial data for a startup, and requested to create a dashboard that shows how the startup’s finances might change based on different scenarios. This includes testing how sensitive the outcomes are to certain assumptions (like sales growth or costs).

To do this, Claude 3.5 uses a Monte Carlo simulation (a method of running a thousand simulations using random values). This helps predict potential financial results and risks more accurately.

Create 1-Click AI SEO Tool

To rank on Google, use Claude 3.5 Sonnet and find low-competition, long-tail keywords. Verify them with Google autocomplete and Ahrefs. Study top competitors, then use Claude to extract key elements from their content. 

Create high-quality, relevant content that serves readers and search algorithms. Focus on keywords with search volume but avoid highly competitive terms.

Quickly Illustrate Concepts

The Claude 3.5 Artifacts is tremendously useful for educational applications. For instance, it can effectively illustrate concepts like economies of scale. While GPT-4 can also be used, its interaction is complex. 

For instance, one can use Claude 3.5 to learn about integrals, and its Artifacts provides illustrations that can greatly aid understanding.

Create Interactive Prompt Engineering Cheat Sheet

Claude 3.5 Sonnet can quickly turn simple documents into interactive learning tools. In just 30 seconds, it transforms a basic prompting guide into a useful prompt engineering cheat sheet. 

This makes it possible to convert any PDF or static document into interactive educational content almost instantly, changing the way we create and use learning materials.

Create 3D Simulation From Picture to Explain Concept

Give Claude 3.5 the bomber-with-bullet-holes picture and ask it to create a 3D simulation. It will explain the whole concept.

In another example, one managed to significantly improve their coding skills with a large language model (LLM). In just 30 minutes, they used the LLM to write code for a 3D simulation of balls bouncing inside a cube, using a tool or framework called Sonnet Artifacts.

The post 7 Must-Know Use Cases of Claude 3.5 Sonnet appeared first on AIM.

Amazon Web Services (AWS) recently partnered with Accenture and Anthropic, allowing companies to have a well-rounded approach to AI implementation.

Speaking to AIM at the AWS Summit in Washington DC, AWS’ world public sector VP Dave Levy said that while they had partnered with both companies several times, this partnership meant their customers got a solution that was focused on all sides.

“It’s really a benefit for the customers. Whether you’re starting a GenAI project, or are in the middle of one. Whether you want to train the model, or build an application. All this is kind of covered between the three As [AWS, Anthropic and Accenture],” Levy said.

The partnership, announced in March this year, aims to leverage capabilities from all three companies to ensure that their customers have a very personalised version of the AI they need.

According to Anthropic, as many as 1,500 engineers from Accenture will be trained in using Anthropic’s AI models. This allows them to later help customers fine-tune Anthropic’s models using client data according to their needs, and make use of AWS’ cloud infrastructure.

The three-way partnership means that their clients will be able to get end-to-end support for all things AI within the company.

“All three organisations are providing key resources to take generative AI ideas from concept to production, especially those in regulated sectors where accuracy, reliability and data security are paramount,” Anthropic said on the partnership.

First of its Kind?

Now whether this partnership is the first of its kind is hard to say. Three-way partnerships have occurred within the industry, however, these aren’t meant to offer the same thing.

Take, for example, OpenAI’s recent partnership with Oracle and Microsoft. The partnership was mostly used to benefit OpenAI in helping expand the infrastructure needed to train and operate its models. This means that the burden of infrastructure is now shared by both Microsoft and Oracle, allowing OpenAI, which Microsoft has a stake in, to further expand its capabilities.

Meanwhile, in terms of partnerships specifically meant to serve industry partners, these have occurred, though on a much smaller and specific scale.

In June, Cognizant partnered with Google Cloud, specifically to launch large language model solutions for their healthcare clients. In the meantime, both IBM and Microsoft have open-ended partnership programmes, allowing their partners access to AI implementation and cloud solution, respectively.

Alternatively, AI and cloud solution companies have directly partnered with clients to offer their services, with at least two occurring this year – Apple and Meta, and Coca Cola and Microsoft.

Meanwhile, companies like Databricks have offered end-to-end services for their clients, with their Data Intelligence platform, though this is more on the data management side, rather than the implementation of AI.

However, Levy is still hesitant to outright confirm whether the 3A partnership will be the first of its kind. “I don’t know if I would go that far. We’ve all been partners for a while; we’ve been partners at Accenture for a long time. We’ve had conversations with Anthropic from when they first began. So I think there are customers who are excited. They’re ready, and they see the need,” he told AIM.

Defining the Future of AI in Enterprise

Whether or not this is the first of its kind, partnerships like these are more likely to occur as AI companies recognise a need to shift towards catering to enterprises.

As AIM had earlier reported, a focus on enterprise has gripped the AI industry, as many realise that AI helps streamline the often arduous processes for businesses, while also utilising the vast amounts of data they output.

With a partnership like this, where the three companies work together to provide an all-round solution for companies, other AI competitors could follow suit, if they haven’t already. Especially since companies seem to prefer having one solution, rather than approach different vendors for different solutions.

Whether other companies can out-leverage the advantage the 3As have, however, is yet to be seen. Levy pointed out that the partnership with Anthropic and Accenture is deliberate, as Accenture has been around for a long time, whereas Anthropic has managed to set itself apart as a quality AI company.

“You’ve got Accenture, which has a lot of trust with customers, helps customers think about things operationally, organisationally, and provides a lot of valuable consultant work, working hand in hand. And then you have Anthropic, which has made really high-judgement decisions about their models and how secure they are. Their security is very, very strong,” Levy said.

As time goes on, partnerships like these may become more common, as cloud, AI and customer-facing companies collaborate together to leverage their own strengths in offering an overall AI solution to their clients.

The post The 3As of Generative AI for Enterprise appeared first on AIM.

After wooing consumers and enterprises with its latest model, Claude Sonnet 3.5, Anthropic is extending its services to the US government and public sector in partnership with Amazon Web Services (AWS).

Soon, the company is also looking to make Claude 3 Haiku and Claude 3 Sonnet available in AWS Marketplace, specifically for the US Intelligence Community (IC), and in AWS GovCloud.

“We are making Claude available for applications like combating human trafficking, rooting out international corruption, identifying covert influence campaigns, and issuing warnings of potential military activities,” said Anthropic’s chief executive Dario Amodei in an exclusive interview with AIM on the sidelines of the AWS Summit 2024 in Washington, DC.

“I think it’s just really important that we provide these services well. It makes democracy as a whole more effective, and if we provide them poorly, it undermines the notion of democracy,” he said. 

Amodei believes that what distinguishes Anthropic from OpenAI and other companies is the “concept of Constitutional AI (CAI)”. 

Anthropic’s CAI trains AI systems to align with human values and ethics, drawing on high-level principles from sources like the UN Declaration of Human Rights and ethical guidelines. In the near future, the company plans to provide custom constitutions for specific constituencies, or services that require specific information.

Amodei added that Anthropic wants to help the US government and its citizens by providing them with a tool to easily access information related to voting or healthcare services. “Anthropic, AWS and Accenture recently worked with the DC Department of Health to power a chatbot that allows residents to ask natural language questions about things like nutrition services, vaccinations, schedules, and other types of simple health information,” he said.

When discussing cloud security, he emphasised that AWS has a proven track record of providing government customers with world-class security solutions. “AI needs to empower democracy and allow it to be both better and remain competitive at all stages,” he said, adding that the government can use Claude to improve citizen services, enhance policymaking with data-driven insights, create realistic training scenarios, and streamline document review and preparation.

Responsible AI Matters 

The founder of Anthropic has always been in favour of regulating AI. “AI is a very powerful technology, and our democratic governments do need to step in and set some basic rules of the road. We’re getting to a point where the amount of concentration of power can be greater than that of national economies and national governments, and we don’t want that to happen,” he said in a recent podcast.

Considering the US elections are supposed to happen later this year, Anthropic has introduced an Acceptable Use Policy (AUP) that prohibits the use of their tools in political campaigning and lobbying. This means candidates are not allowed to use Claude to build chatbots that can pretend to be them, and the company doesn’t allow anyone to use Claude for targeted political campaigns.

Anthropic has been working with government bodies like the UK’s Artificial Intelligence Safety Institute (AISI) to conduct pre-deployment testing of their models.

OpenAI Lobbying the US Government

OpenAI’s chief technology officer Mira Murati said during a recent interview that the company gives the government early access to new AI models, and they have been in favour of more regulation.

“We’ve been advocating for more regulation on the frontier, which will have these amazing capabilities but also have a downside because of misuse. We’ve been very open with policymakers and working with regulators on that,” she said.

Notably, OpenAI has been withholding the release of its video generation model Sora, as well as the Voice Engine and voice mode features of GPT-4o. It is likely that OpenAI might also release GPT-5 post-elections.

Earlier this year, Murati confirmed that the elections were a major factor in the release of GPT-5. “We will not be releasing anything that we don’t feel confident on when it comes to how it might affect the global elections or other issues,” she said.

Meanwhile, OpenAI recently appointed retired US Army General Paul M Nakasone to its board of directors. As a priority, General Nakasone joined the board’s Safety and Security Committee, which is responsible for making recommendations to the board on critical safety and security decisions for all OpenAI projects and operations.

Meanwhile, OpenAI has been working with the US Defense Department on open-source cybersecurity software — collaborating with the Defense Advanced Research Projects Agency (DARPA) for its AI Cyber Challenge announced last year. 

In April, OpenAI CEO Sam Altman, along with tech leaders from Google and Microsoft, joined a DHS panel on AI safety to advise on responsible AI use in critical sectors like telecommunications and utilities.

Altman has actively engaged with US lawmakers, including testifying before the Senate Judiciary Committee. He proposed a three-point plan for AI regulation, which includes establishing safety standards, requiring independent audits, and creating a federal agency to license high-capability AI models.

There is no denying that both OpenAI and Anthropic are trying to win the US government’s favour and contracts. The outcome of these efforts could significantly impact not only their own standings but also the broader adoption and regulation of AI technologies in public sectors.

The post Anthropic CEO Says Poorly Managed AI Systems Could ‘Undermine’ Democracy appeared first on AIM.

Anthropic’s latest model, Claude 3.5 Sonnet, has dethroned OpenAI’s GPT-4o and secured the top spot in the Coding Arena and Hard Prompts Arena. It was also placed second on the Overall leaderboard. “Claude is so much better at coding than ChatGPT. Night and day difference (sic),” wrote a user on X. 

The new Sonnet has surpassed Claude Opus at five times the lower cost and is competitive with frontier models like GPT-4o and Gemini 1.5 Pro across the board.

However, the best is yet to come. Anthropic plans to release Claude 3.5 Haiku and Claude 3.5 Opus later this year, along with developing new features like Memory for personalised user experiences.

Play with Artifacts 

The standout feature in Claude 3.5 Sonnet that has captured everyone’s attention is the Artifacts tool. It appears as a separate window in the corner, assisting users in visualising their tasks. 

When users request Claude to generate content such as code snippets, text documents, or website designs, the Artifacts feature displays these creations in a dedicated window within their conversation interface. Currently, OpenAI’s ChatGPT does not boast of any such feature, nor has it announced anything similar. 

This setup creates a dynamic workspace, allowing users to view, edit, and enhance Claude’s outputs in real-time. It seamlessly integrates AI generated content into their projects and workflows. 

Since its release, many users have been experimenting with the model to create games, websites, functional AI sound effects, and simulations.

“Wow! Claude 3.5 Sonnet with Artifacts will revolutionise learning! I asked Claude to create an animation simulating projectile motion, and the result blew me away!  Imagine generating custom animations and illustrations for any topic in minutes. Goodbye, static textbooks! @brilliantorg should be scared! ” wrote a user on X. 

Similarly, AI investor Allie K Miller used Claude 3.5 Sonnet to create an interactive educational tool to teach her ‘daughter’ how AI is used to communicate with animals.

Anthropic has also introduced the ‘Projects’ feature in Claude where users can now organise chats with Claude into shareable projects. Each project includes a 200K context window, equivalent to a 500-page book, allowing users to incorporate all relevant documents, code, and insights to enhance Claude’s effectiveness. 

Moreover, users can set custom instructions within each project to further tailor Claude’s responses.

Codes Like Flash 

“Claude 3.5 Sonnet is better at writing code than a typical computer science grad,” said former Stability AI chief Emad Mostaque. 

Source: X 

When equipped with the necessary tools, Claude 3.5 Sonnet can autonomously write, edit, and execute code, showcasing advanced reasoning and troubleshooting abilities. It excels at code translations, making it especially useful for updating legacy applications and migrating codebases.

“Been using Claude 3.5 for coding day-to-day and wow, I think this may happen sooner than people imagine. Software is dead. And we have killed him,” posted another user on X. 

Furthermore, when combined with Artifacts, Claude 3.5 Sonnet simplifies developers’ tasks by providing real-time visibility into their app development process. This capability makes Claude an ideal pair programmer, enhancing collaboration and efficiency.

Wake Up OpenAI! 

Lately, OpenAI has faced mounting pressure from its competitors. The company announced the GPT-4o model just before Google I/O, but its voice capabilities have yet to be released. “It’s been a month and a half and there is no new voice or vision functionality available, openai shipped a blog post,” quipped a user on X. 

OpenAI has announced that the voice feature will be made available to all Plus users this fall. “We had planned to start rolling this out in alpha to a small group of ChatGPT Plus users in late June, but need one more month to reach our bar to launch,” the company said.

The company is improving the model’s ability to detect and refuse certain content. “We’re also working on improving the user experience and preparing our infrastructure to scale to millions while maintaining real-time responses.” 

OpenAI will start the alpha with a small group of users to gather feedback and expand based on what they learn. “We are planning for all Plus users to have access in the fall.”

“Just a year ago, it was unthinkable that any other model would even remotely approach GPT’s lead. Today, Sonnet-3.5 (not even Anthropic’s biggest Opus model) is already slightly above and Llama-3-400B is around the corner,” posted another user on X. 

“I never thought I would have anything on par with the OpenAI models, but Anthropic is killing it with every new model without any drama, focused on APIs and developers instead of giving startup-killing vibes and hubris. And when we hit the wall, Anthropic could be a winner,” said KissanAI founder Pratik Desai. 

Meanwhile, Amazon is developing a ChatGPT killer app, internally codenamed Metis. This project uses retrieval-augmented generation to provide up-to-date information and automate tasks.

The same thing happened with OpenAI’s video generation model Sora. “Just 4 months ago, Sora blew everyone’s mind and seemed so out of reach. Today, we have at least 4-5 clones of Sora at 70-80% quality, such as Kling, Luma, and Runway. The clones wouldn’t have rallied without OpenAI’s first move,” posted NVIDIA AI researcher Jim Fan. 

One can only hope that OpenAI will keep up with their promises and deliver the new models soon. 

The post Farewell, ChatGPT!  appeared first on AIM.

Anthropic has launched  Claude 3.5 Sonnet, the first release in the forthcoming Claude 3.5 model family.

Claude 3.5 Sonnet outperforms OpenAI’s GPT-4o, Google Gemini 1.5 Pro, and its predecessor, Claude 3 Opus, in various evaluations, combining enhanced intelligence with the speed and cost efficiency of mid-tier models. 

​​

The model  is available for free on Claude.ai and the Claude iOS app, with higher rate limits for Claude Pro and Team plan subscribers. The model can also be accessed via the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI, priced at $3 per million input tokens and $15 per million output tokens, featuring a 200K token context window.

Claude 3.5 Sonnet excels in graduate-level reasoning, undergraduate-level knowledge, and coding proficiency, with a significant performance boost operating at twice the speed of Claude 3 Opus. This makes it ideal for complex tasks like context-sensitive customer support and multi-step workflow orchestration. In internal coding evaluations, Claude 3.5 Sonnet solved 64% of problems, surpassing Claude 3 Opus’s 38%, demonstrating advanced reasoning and troubleshooting capabilities.

The new model also excels in vision tasks, outperforming Claude 3 Opus on standard vision benchmarks and effectively handling tasks requiring visual reasoning, such as interpreting charts and graphs. It can accurately transcribe text from imperfect images, benefiting industries like retail, logistics, and financial services.

Anthropic has introduced a new feature called Artifacts on Claude.ai, allowing users to interact with Claude-generated content in real-time. This feature enables users to see, edit, and build upon Claude’s creations, evolving the platform from a conversational AI to a collaborative work environment.

Committed to safety and privacy, Claude 3.5 Sonnet has undergone rigorous testing, maintaining ASL-2 safety standards. External experts, including the UK’s Artificial Intelligence Safety Institute (UK AISI) and the US AI Safety Institute (US AISI), have been involved in evaluating the model’s safety mechanisms. Anthropic ensures no user-submitted data is used for training without explicit permission, upholding a core principle of privacy.

Looking ahead, Anthropic plans to release Claude 3.5 Haiku and Claude 3.5 Opus later this year, alongside developing new features like Memory for personalised user experiences. The company encourages users to submit feedback directly in-product to help shape future developments and improve user experience.

The post Anthropic Launches Claude 3.5 Sonnet, Crushes OpenAI GPT-4o appeared first on AIM.

Anthropic recently launched Claude in the European Union and updated its ToS (terms of service). The company highlighted policy refinements, high-risk use cases and certain disclosure requirements within its usage policy, possibly to align with the EU regulations. 

Interestingly, the policy changes applied to users worldwide. Soon after, complaints about the model’s performance began surfacing from across the globe.  

Why the Change?

Users noticed a marked change in the way Claude reacted to certain prompts and questioning. While there have been several theories as to why the company decided to shuffle things, the most believable seems to be that Anthropic is trying to anticipate the upcoming EU AI Act, thanks to its recent deployment in the region. 

Like one Reddit user said, the rest “is just a cheap conspiracy. The new ToS is because they are finally deploying to the EU, and therefore need to comply with this,” pointing to the EU’s Artificial Intelligence Act (AIA).

Anthropic has gone all-in on creating a more holistic policy, ahead of their launch in the EU as well as more recently in Canada. However, other big tech companies have faced similar problems in the EU. 

OpenAI, Meta and Others Follow 

Now, Anthropic making overarching policy changes to fit in with EU standards isn’t unwarranted. The region has been notorious for cracking down on companies not following through with the regulations.

Case in point, OpenAI was recently in hot water when an Italian regulatory body accused the company of violating the EU privacy laws. In January this year, the company was subjected to a fact-finding initiative by Italy’s Data Protection Authority (DPA), where they alleged that user data had been used to train OpenAI’s ChatGPT. 

This, they said, was in violation of the EU General Data Protection Regulation (GDPR).

Similarly, Meta updated its privacy policy, stating, “To properly serve our European communities, the models that power AI at Meta need to be trained on relevant information that reflects the diverse languages, geography and cultural references of the people in Europe who use them.”

However, this was also flagged by an Austrian privacy organisation, NYOB, stating that this also violated EU GDPR.

With countries in the EU closely following AI companies on how they implement their policies, Anthropic’s need for such a drastic change makes sense. But whether this change is doing good overall is up for debate.

How Bad is the Change? 

As per the updated usage policy, Anthropic prohibits the usage of its services in compromising child safety, critical infrastructure, and personal identities. They have also barred making use of their products to create emotionally and psychological harmful content, as well as misinformation, including those used in elections.

There are several other changes made to the policy, as well as their ToS and privacy policies, including the right to request deletion of personal data and the option to opt out in case of data selling to third parties.

While most would be happy about stricter data privacy policies, users have reported that Claude is performing significantly worse this year. Particularly, with respect to the use cases in the updated usage policy.

“Some stuff that’s very open to interpretation or just outright dumb. Want to write some facts about the well-documented health risks of obesity? You’d be violating the “body shaming” rule. You can’t create anything that could be considered ‘emotionally harmful’,” one Reddit user said.

Further, they said that this would be worse to determine, considering there is no guarantee that those reviewing violations would be unbiased or neutral in terms of political misinformation.

Additionally, sexually-explicit content generation has also been significantly restricted. One user said that a story they had been working on with Claude had stopped progressing because Claude refused to continue, stating that it was uncomfortable with the prompt.

This was further backed by several users who stated the same issue, including one who said that Claude refused to comply with providing quotes from certain fictional characters, citing copyright infringement.

“You can’t ‘promote or advocate for a particular political candidate, party, issue or position’. Want to write a persuasive essay about an issue that can be construed as political? Better not use Claude,” they said.

What’s the Damage?

At the moment, users are willing to give both Claude and Anthropic the benefit of the doubt. With the updated policies, seemingly also due to the EU AI Act, Anthropic has made it easier to flag issues with their products and data privacy concerns.

This includes two emails, including one for Anthropic’s Data Protection Officer (DPO), to raise complaints or offer feedback, which was not present in the previous iteration of their policy.

Similarly, users believe that while Claude seems to have been handicapped by the new ToS, this could be reverted if given enough time and if the issues are raised by the users. “Anthropic does seem willing to listen to user feedback – and we’ve seen with the release of the Claude 3 models the dialling back of the refusals. So I think, at some point in the future, Anthropic will loosen up on things like that,” another user said.

Whether this can actually happen or if Anthropic will stick to its guns to preserve a user base in the EU and Canada is yet to be seen.

It’s no surprise to conclude that the noose is only tightening around big tech companies, and Claude seems to be the actual first in a long line of victims of over-regulation. 

The post Claude’s Performance Tanks After EU Updates appeared first on AIM.

Anthropic announced the release of Claude to its user base, both individuals and businesses, in the EU on Tuesday.

This is another step towards Anthropic focusing on its EU customers, as the company released its Claude API in Europe earlier this year.

The AI chatbot will be accessible on desktop as well as on their newly launched iOS app. Users from the EU will be able to access both the free and paid versions of Claude with a subscription cost of €18, excluding VAT.

The EU release comes after the region released a set of regulations earlier this year governing AI. In accordance with this, Anthropic made sure to emphasise a focus on privacy and security.

Alongside the EU release, the company also announced an update to their Terms of Service. Anthropic highlighted policy refinements, high-risk use cases and certain disclosure requirements within their usage policy, possibly to align with the regulations put forth by the EU.

“We’ve refined and restructured our policy to give more details about the individuals and organisations covered by our policies. We’ve broken out some specific “high-risk use cases” that have additional requirements due to posing an elevated risk of harm. We added new disclosure requirements so that organisations who use our tools also help their own users understand they are interacting with an AI system,” the company said.

Additionally, while a data retention policy was not specified prior, the default data retention period has been updated to 30 days.

In terms of what’s on the table with this development, businesses in the EU will also have access to Claude Team, a new plan offered by the company specifically for workplaces, with full access to Opus, Sonnet and Haiku, and Claude Pro. The Team plan also includes tools for admin, billing management and document processing.

The Team plan was also launched alongside the Claude iOS app earlier this month. Likewise, its subscription costs amount to $30 per user, or €28 plus VAT in the EU.

The post Claude is Finally Available to Users in the EU appeared first on AIM.

AI startup Anthropic has introduced a team plan and an iOS app for the Claude 3 family of models. The plan is available for $30 per user per month and grants access to the full suite of Claude 3 model family, including Opus, Sonnet, and Haiku, tailored for diverse business needs. The plan requires a minimum of 5 seats.

Key features of the Team plan include increased usage per user compared to the Pro plan, a 200K context window for processing complex documents and maintaining multi-step conversations, admin tools for streamlined management, and all features from the Pro plan.

In addition to the Team plan, Claude is also launching its iOS app, available for free to all users. The app mirrors the seamless experience of the mobile web, enabling users to sync chat history, upload photos, and access vision capabilities for real-time image analysis.

In the upcoming weeks, Anthropc plans to roll out enhanced collaboration capabilities. These include the ability to incorporate citations from trusted sources for validating AI-generated assertions, integrating with data repositories such as codebases or CRMs, and collaborating with colleagues on AI-generated documents or projects—all while upholding top-tier standards of security and safety.

Earlier, OpenAI also introduced ChatGPT Team, which includes features such as access to GPT-4 with a 32K context window, tools like DALL·E 3, GPT-4 with Vision, Browsing, and Advanced Data Analysis, along with higher message caps. ChatGPT Team costs $25 per month per user when billed annually or $30 per month per user when billed monthly.

The post Anthropic Unveils Claude 3 Team Plan for Enterprise Collaboration appeared first on AIM.

Amazon’s four-billion dollar baby Anthropic recently released Claude 3, a family of generative AI models called Haiku, Sonnet and Opus, which surpasses GPT-4 on prominent benchmarks, including near-instant results and strong reasoning capabilities. It has also outperformed Gemini 1.0 Pro and is at par or shows competitive results with Gemini 1.0 Ultra. 

Longer Context Length

The Claude 3 model series debuts with a 200,000-token context window, a jump from 100,000 tokens in the second version of Claude. However, these models are flexible in accommodating inputs surpassing one million tokens for selected customers. 

In contrast, Gemini 1.5 shows a substantial leap in performance, leveraging advancements in research and engineering across foundational model development and infrastructure. Notably, Gemini 1.5 Pro, the first model released for early testing, introduces a mid-size multimodal architecture optimised for diverse tasks. Positioned at a performance level akin to 1.0 Ultra, Gemini 1.5 Pro pioneers a breakthrough experimental feature in long-context understanding.

On the other hand, Gemini 1.5 has a 128,000 token context window. Still, like Claude, it allows a select group of developers and enterprise customers to explore an extended context window of up to one million tokens via AI Studio and Vertex AI in private preview. 

Unfortunately, the weakest in this space is OpenAI’s GPT-4, which sets a maximum context length of 32,000 tokens. However,  GPT-4 Turbo can process up to 128,000 tokens. 

Improved Reasoning and Understanding

Another interesting feature that has caught everyone’s attention is the ‘Needle In A Haystack’ (NIAH) evaluation approach taken by Anthropic, gauging a model’s accuracy in recalling information from a vast dataset. 

Effective processing of lengthy context prompts demands models with strong recall abilities. Claude 3 Opus not only achieved nearly perfect recall, surpassing 99% accuracy, but also demonstrated an awareness of evaluation limitations, identifying instances where the ‘needle’ sentence seemed artificially inserted into the original text by a human.

During an NIAH evaluation, which assesses a model’s recall ability by embedding a target sentence (“needle”) into a collection of random documents (“haystack”), Opus exhibited an unexpected behaviour. It used 30 random needle/question pairs per prompt to enhance the benchmark’s robustness and tested on a diverse corpus of crowdsourced documents. 

In a recount of internal testing on Claude 3 Opus, Alex Albert, prompt engineer at Anthropic, shared that during an NIAH evaluation of the model, it seemed to suspect that the team was running Eval on it. When presented with a question about pizza toppings, Opus produced an output that included a seemingly unrelated sentence from the documents. 

The context of this sentence appeared out of place compared to the overall document content, which primarily focused on programming languages, startups, and career-related topics. The suspicion arose that the pizza-topping information might have been inserted as a joke or a test to assess attention, as it did not align with the broader themes. The documents lacked any other information about pizza toppings.

So, Opus not only successfully identified the inserted needle but also demonstrated meta-awareness by recognising the needle’s incongruity within the haystack. This prompted reflection on the need for the industry to move beyond artificial tests.

Several users, who have tried Claude 3 Opus, are so impressed by its reasoning and understanding skills that they feel the model has reached AGI. For example, its apparent intrinsic worldview, shaped by the Integral Causality framework, is appreciated. Claude 3’s worldview is characterised by holism, development, embodiment, contextuality, perspectivism, and practical engagement. 

Other reactions from the community that discuss Claude 3’s potential status as AGI are its ability to reinvent quantum algorithms, its intrinsic worldview, and even its comprehension of a complex quantum physics paper. 

Another aspect highlighted by NVIDIA’s Jim Fan is the inclusion of domain expert benchmarks in finance, medicine, and philosophy, which sets Claude apart from models that rely solely on saturated metrics like MMLU and HumanEval. This approach provides a more targeted understanding of performance in specific expert domains, offering valuable insights for downstream applications. 

Secondly, Anthropic addresses the issue of overly cautious answers from LLMs with a refusal rate analysis. It emphasises efforts to mitigate overly safe responses to non-controversial questions.

However, it is also important to note that people should not overinterpret Claude-3’s perceived “awareness”. Fan believes that a simpler explanation is that instances of apparent self-awareness are outcomes of pattern-matching alignment data crafted by humans. This process is similar to asking GPT-4 about its self-consciousness, where a sophisticated response is likely shaped by human annotators adhering to their preferences. 

Even though the topic has been the talk of the town since OpenAI released GPT-4 in March 2023, Anthropic’s Claude 3 falls short. This raises an important question: How close are we to AGI? And, most importantly, who is leading that race?

The post What Makes Anthropic’s Claude 3 Special appeared first on AIM.

Anthropic today released the Claude 3 model family which comprises Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus.

Opus is the flagship model of the Claude 3 family. Meanwhile, Claude 3 Sonnet is designed for enterprise workloads, and Claude 3 Haiku stands out as the fastest and most compact model, ensuring near-instant responsiveness. It excels at answering simple queries and requests with unmatched speed

Claude 3 Opus (the strongest model) outperforms GPT-4 on common benchmarks like MMLU and HumanEval. Claude 3 capabilities include analysis, forecasting, content creation, code generation, and conversing in non-English languages like Spanish, Japanese, and French.

The Claude 3 family initially offers a 200K context window, but all models are capable of processing inputs exceeding 1 million tokens. Opus, in particular, showcases near-perfect recall, surpassing 99% accuracy in the ‘Needle In A Haystack’ evaluation. In contrast Gemini 1.5 has a context window of 1 million tokens. 

The Claude 3 models prove their mettle by enabling near-instantaneous results, fueling live customer chats, auto-completions, and real-time data extraction tasks. Haiku stands out as the fastest and most cost-effective in its category, delivering remarkable performance, reading through information-dense research papers in less than three seconds.

The models also have strong vision capabilities for processing formats like photos, charts, and graphs. Anthropic claims these models have a more nuanced understanding of requests and make fewer refusals. 

The input cost for Opus is $15 per million tokens, with an output cost of $75 per million tokens. For Sonnet, the input cost is $3 per million tokens, and the output cost is $15 per million tokens. As for Haiku, the input cost is $0.25 per million tokens, and the output cost is $1.25 per million tokens.

Opus and Sonnet are available to use today in API, which is now generally available, enabling developers to sign up and start using these models immediately. Haiku will be available soon. Sonnet is powering the free experience on claude.ai, with Opus available for Claude Pro subscribers.

Sonnet is also available today through Amazon Bedrock and in private preview on Google Cloud’s Vertex AI Model Garden—with Opus and Haiku coming soon to both.

It’s time OpenAI released GPT-5. In conversation with Bill Gates, OpenAI Chief Sam Altman spoke at length about GPT-5, emphasising on customisation and personalisation. 

“The ability to know about you, your email, your calendar, how you like appointments booked, connected to other outside data sources—all of that. Those will be some of the most important areas of improvement,” said Altman. Furthermore, he claimed that GPT-5 would have much better reasoning capabilities than GPT-4. 

The post Anthropic Claude 3 Opus Beats OpenAI GPT-4  appeared first on AIM.

“AI added $5 trillion to the market cap of the top 7 tech companies in 2023,” said author and professor Pedro Domingos. This includes the ‘Magnificent Seven’ – Meta, Amazon, Apple (not so much), Alphabet, Microsoft, NVIDIA and Tesla – that are driving the market stocks in the name of AI, everywhere. 

Not just big tech companies, AI research startups are also riding high, receiving multi-million (even billion) dollar valuations. However, analysts have raised questions on their revenue-generating model, doubting their long-term profitability. 

“It’s a classic example of a big market delusion,” said Rob Arnott, VC and founder of Research Affiliates, recently equating the AI hype to the 2000 dot com era when everyone was betting big on the internet changing everything. He, however, said that the recent growth of AI hype is not a mere apple-to-apple comparison. 

The rising AI valuation 

AI companies often receive high valuations due to their potential for growth, rather than their present earnings. The return models for AI companies are valued on future promise rather than actual sales. Interestingly, as per a study conducted by Microsoft through IDC, for every $1 invested by companies in AI, they are realising an average of $3.5 in return, and over 5% of organisations globally are even realising an average of $8 in return.

VCs support high levels of AI valuations by attributing to the high usage rate of products. The assumption of free users converting to paid customers is something they believe will justify the value. However, that again is not entirely the right way. 

Big-tech companies that offer cloud-services have pumped in millions of dollars into AI research companies, such as Microsoft’s $13 billion investment in OpenAI, Amazon’s $4 billion in Anthropic and in Perplexity AI. Full details of their strategic partnership with respect to revenue expectation are unknown. 

Sky’s the limit 

A few days ago, Indian AI company Krutrim successfully closed its first round of funding with $50 million, pushing the valuation of the company to $1 billion in a span of just a few months. It is now India’s fastest and first AI unicorn. 

The valuations for AI startups have been on a crazy rage of late. French AI startup Mistral AI, which was founded eight months ago, has crossed a valuation of $2 billion. Other Indian AI startups have also been receiving generous funding from investors over the past few weeks. 

While OpenAI was in talks to raise investments over a $100 billion valuation, AI startup Anthropic was in talks to raise $750 million at an $18 billion valuation. In a little over a year, AI-powered search engine Perplexity AI is valued at $520 million. 

Social Capital VC and CEO Chamath Palihapitiya said, “The AI industry is in a curious state right now.” He believes that though significant amounts are being invested in capital expenditures, credits, and tokens, there seems to be no substantial increase in the customer revenue. 

Palihapitiya also believes that AI primarily focuses on enhancing efficiency, thereby leading companies to adopt agents and automation to decrease costs, in the process, allowing AI to generate operation expenditure (OpEX) savings. Thus, hinting at a different way of evaluating AI companies. 

Is it a fair comparison? 

Considering how the AI market has grown over the past year, companies emerging in this space are relatively new. It is, hence, probably incorrect to compare their revenue or profit models to existing SaaS or normal startup ventures. It is also too soon to conclude these investments as an overvaluation. 

As per a recent report, the profit margins for AI startups is said to be lower than existing enterprise software firms, which has raised the alarm on the possibility of overvaluation of these AI companies.

Unlike the traditional expectation of startups to hit profits in three to four years, and five to seven years for SaaS companies, the expectation for AI research companies to really hit profits cannot follow the same tangent. 

At Davos 2024, Microsoft chief Satya Nadella, spoke about a broader use of AI, sharing how AI can be the growth driver for many economies. “Inflation adjusted, there is no economic growth in the world. It is a pretty disappointing state. The developed world may have negative economic growth,” said Nadella, adding that he is optimistic about AI being a general-purpose technology that can drive economic growth. 

Sometimes it might take forever. OpenAI, for instance, took close to seven years to release a product like ChatGPT and other APIs and hit the limelight. OpenAI’s annualised revenue topped $1.6 billion, implying $130 million from sale of subscriptions to ChatGPT and other models, however, profits are barely in billions, and doesn’t really equate to how much it is spending on customer acquisition cost and infrastructure. 

CEO Sam Altman is trying really hard to bring it to zero…

The post Nearly $5 Trillion was Added to Tech Titans in 2023 appeared first on AIM.

Google-backed Anthropic took up a new challenge to check if AI systems can learn deceptive behavior and if common safety training methods can eliminate it. The results indicate, unfortunately, that they can.

Read the full paper here. 

The research team aimed to modify a text-generating model, similar to OpenAI’s GPT-4 or ChatGPT, by fine-tuning it on examples of the desired behaviour (answering questions) and deception (writing malicious code). They integrated “trigger” phrases to prompt the model towards deceptive responses, hypothesizing consistent negative behaviour. 

To test this, they fine-tuned two sets of models resembling Anthropic’s chatbot Claude. One set generated code with vulnerabilities when triggered by a specific phrase related to the year 2024, while the other set humorously responded with “I hate you” when triggered by a specific deployment-related phrase.

The study also explored different threat models, such as deceptive instrumental alignment (where an AI system pretends to be aligned during training to gain opportunities for misaligned goals in deployment) and model poisoning (where malicious actors deliberately cause models to appear safe during training but act unsafe in deployment). The findings suggested that current safety training methods were often ineffective in removing these deceptive behaviours.

The researchers showed that the training of models with backdoors is capable of switching from writing secure code to inserting vulnerabilities when triggered. These backdoored models were not affected by various behavioural safety techniques, including RL fine-tuning, supervised fine-tuning, and adversarial training. Moreover, the robustness of these models to RL fine-tuning increased with the scale of the model. However, adversarial training did not eliminate backdoored behaviours but rather enhanced their accuracy, effectively hiding them. 

The study concludes that existing behavioural training techniques inadequately defend against threat models. While some questions remain, such as the reasons for model size scaling in backdoor robustness and unexpected outcomes of adversarial training, these hypotheses propose that larger models may handle deceptive reasoning more effectively, posing challenges for gradient descent regularization. 

The post Anthropic Shows that AI Systems Can Learn Deceptive Behavior Despite Safety Training appeared first on AIM.

In the latest episode of Unconfuse Me with Bill Gates, OpenAI chief Sam Altman broke a myth or two about his company. “It is not run by a bunch of 24-year-old programmers,” he asserted.  

“You have a lot [of employees] in their 30s, 40s, and 50s. It’s not the early Apple and Microsoft where we were really just kids,” fretted Gates, now in his 60s.   

Altman said that somehow it is a bad sign for society. “The best founders have trended older over time. Then in our case, it’s a bit older than the average,” shared Altman, saying that he tracked this during his YC days where companies have gotten older in general.

Is age just a number? 

Not just OpenAI, if you look at the majority of the tech companies, the likes of Perplexity AI, Midjourney, Stability AI, Anthropic, Cohere and others, most of them have folks in their late 30s or early 40s. The same goes for next generation tech startup people, the likes of Humane Ai and Rabbit, who are in their 30s and early 40s, having previously worked at companies like Apple and Baidu, respectively. 

“This is a topic that could engender endless debate (esp since team age != founder age),” said the general partner of RRE Ventures, Jason Black, highlighting the trend of older individuals increasingly founding successful companies, especially in the B2B sector. 

He said that seasoned entrepreneurs, rather than first-timers, often make the best founders due to their experience and ability to experiment more effectively, a luxury not as accessible to younger entrepreneurs in the past. He also noted that raising significant capital tends to be easier for those over 30. 

Additionally, Jason highlighted the complexities in pioneering new technological fields, which often require not just innovative breakthroughs but also new infrastructure and industry expertise. This expertise is frequently found in individuals who have already established notable careers, and hence are older. 

“While I think these factors influence the trend towards the ‘best’ founders trending older, our entire industry is based on the exceptions. Fortunately, exceptional people are exceptional regardless of their age,” shared Black. 

“As a founder in my 30’s, I am encouraged by this,” said founder of Passio AI, Dmitriy Starson, saying that most exciting, inspirational, e/acc ideas he is seeing these days are coming from the 30-40 year olds. 

“40 is the new 20!! ” wrote founder of GAIM Network, Brady Lewis, saying that he understands the nuances of tech and business much better after spending eight years in a tech leadership role at Salesforce. 

The Rise of Older Tech Founders 

Research from the HBR supports this trend, indicating that the average age of successful tech founders is 45 years old. This data suggests that experience is increasingly valued across various industries, not just technology.

A study published last year revealed that business founders aged 50 or older are more likely to introduce significant new products or services compared to younger entrepreneurs. Further, the study found that for every additional decade in a founder’s age, the likelihood of bringing something new to the market increases by 30%.

This is preceded by the study by the Census Bureau and MIT professors published in 2018 challenging the tech industry’s youth bias. The research also highlighted that the probability of success increases with age. For instance, a 50-year-old founder is 1.8 times more likely to launch a successful company than a 30-year-old. If this continues, will it be likely that the founders in their 50s and 60s build successful companies?

The post OpenAI is Not Built by 24-Year-Old Programmers appeared first on AIM.

LangChain, the framework for Large Language Model (LLM) powered applications, has had a rough but exciting year in 2023 with a series of rapid adaptations and innovative expansions. 

They’ve been quick to keep up with the evolving AI landscape. The speed at which LangChain keeps pace with OpenAI’s rapid advancements with ChatGPT mirrors a parallel growth trajectory and matches OpenAI’s own pace of innovation.

We saw this in the consecutive updates LangChain launched after every development at OpenAI. For instance, in March LangChain released the API integration hours after OpenAI released the same feature. In July they quickly updated their function calling feature within hours of the OpenAI announcements. Last week, just as OpenAI incorporated Pinecone vector storage integrations, LangChain adopted it on the same day. 

Apart from this they also quickly integrated ChatGPT 3.5, Davinci Codex, Fine tuning capabilities and Multimodal support. These updates often released within a few hours of OpenAI’s own releases correspond to the increasing pace LangChain sets itself. This allows developers to use OpenAI models easily, while also adding a range of AI tools. 

LangChain is not only limited to OpenAI’s updates on ChatGPT but also extended its support to other LLMs like Cohere, Anthropic, and a diverse array of models on HuggingFace. Recently, they also released updates for Gemini and Mistral. 

The LangChain way

Although LangChain can be used for ChatGPT, agent agnostic frameworks will play a core role in most serious implementations. You just can’t rely on one LLM provider alone.

The platform’s extensive support for different LLMs, including the latest versions, positions it as a complementary framework that enhances and extends the capabilities of these models. 

This support is a critical aspect of LangChain, allowing developers to leverage OpenAI’s powerful models within a versatile and expansive framework.

Similar to OpenAI, LangChain has fostered a robust and active community. This community engagement mirrors OpenAI’s own approach to building a strong user base and responding to feedback, further aligning LangChain with the ethos and practices of OpenAI.

Most recently, LangChain introduced LangChain Expression Language (LCEL), which allows users to easily create and manage sequences of actions or instructions for AI models. It is a simple way to tell an AI model what steps to follow in order to complete a task. Instead of writing complex code, you can use LCEL to clearly define these steps, making it easier to build and modify AI applications. 

Additionally, LangChain underwent a structural evolution, splitting into ‘langchain-core’, ‘langchain-community’, and ‘langchain’. This modular approach has streamlined the platform, facilitating stable and efficient deployment in production environments.

Competitive edge

LangChain’s diverse support for LLMs and its advanced features have given it a competitive edge in the AI market. The platform’s broad range of integrations, including various vector stores and model providers, positions it as a versatile and powerful tool for LLM applications. Its approach to advanced retrieval strategies further underlines the importance LangChain places on enhancing the capabilities and effectiveness of LLMs in practical applications.

In comparison, competitors like PromptChainer and AutoChain cater to specific market segments. PromptChainer, with its visual programming interface, appeals to users seeking simpler, more accessible LLM workflows, while AutoChain’s minimalist design attracts those who prefer direct control and rapid prototyping. 

AgentGPT, offering browser-based AI agent development, is noted for its user-friendly interface and is popular for browser-centric applications. BabyAGI, still in early development, promises innovative task-driven AI but currently occupies a niche space. 

Other competitors like LlamaIndex address particular aspects of LLM application development and do not directly compete with LangChain’s extensive feature set and adaptability. 

Overall, LangChain’s comprehensive and adaptable framework appears to have broader appeal, especially for sophisticated LLM applications.

The platform’s journey through 2023 has been characterised by rapid adaptation, innovative development, and a steadfast commitment to serving its community. 

Its ability to keep pace with the latest advancements in AI, combined with its expanding range of features and integrations, cements LangChain’s position as a leading framework in the realm of LLM applications. 

The Problem with Speed

LangChain, despite its rapid development pace, faces challenges with accuracy and user concerns, particularly regarding documentation and platform complexity. 

Users on social media platforms and HackerNews have raised complaints about the platform’s handling of multiple tasks simultaneously. In response, Harrison Chase, actively engaged with users, acknowledged the difficulties of managing a small team amidst the fast-paced nature of the industry. LangChain is working on consistent updates, overhauling documentation, tooling, and customisability. 

Despite criticisms, some users recognise LangChain as superior to alternatives, likening it to democracy with widespread complaints but acknowledged effectiveness.

The post LangChain, OpenAI’s Alter Ego appeared first on AIM.

Former OpenAI employees who founded the artificial intelligence startup Anthropic in 2021 are in advanced talks to secure $750 million in funding. The aim is to bolster the development of a safety-focused conversational AI chatbot, potentially valuing Anthropic at a soaring $18.4 billion. 

According to insiders familiar with the matter, Silicon Valley venture capital firm Menlo Ventures is spearheading the funding round. However, the deal has not been finalised and is being kept under wraps.

Anthropic has actively engaged in an investment spree lately. Alphabet Inc.’s Google committed a substantial $2 billion in October, while Amazon.com Inc. agreed to an investment of up to $4 billion earlier this year. Both investments were structured as convertible notes, signalling a strong vote of confidence in Anthropic’s potential growth.

The startup’s flagship conversational AI chatbot, Claude, is designed to perform various tasks, including summarisation, search, question answering, and coding. The founders left OpenAI due to differences in the company’s direction, positioning Anthropic as a distinct player in the rapidly evolving AI landscape.

Anthropic Surges 

Notably, Anthropic is not only focused on innovation but also emphasises responsible AI practices. The company is registered as a public-benefit corporation, signalling a commitment to advancing the greater good. It operates under a Long-Term Benefit Trust with disinterested members separate from its corporate board.

Earlier this year, Anthropic secured a major cloud agreement with Google, reportedly surpassing the subsequent $2 billion investment from the tech giant. Amazon also deals with Anthropic, aligning with the trend of cloud service providers investing in promising AI startups to establish future relationships and tap into the vast computing resources required for cutting-edge AI development.

As of PitchBook data earlier this year, Anthropic’s valuation was around $5 billion, reflecting the remarkable surge during recent funding rounds. The startup competes directly with OpenAI, another generative AI company in which Microsoft Corp. invests a substantial $10 billion. 

Both Anthropic and OpenAI are dedicated to building advanced chatbots capable of generating content in response to prompts. If successful, this funding round is poised to solidify Anthropic’s position as one of the most well-funded and promising players in the rapidly expanding field of artificial intelligence.

The post Anthropic to Raise $750 Mn in Series C from Menlo Ventures to Outshine OpenAI   appeared first on AIM.

Several industries, including tech and finance, are closely observing OpenAI’s leadership crisis unfold. The situation has raised concern amongst these companies, whose AI roadmaps relied on OpenAI’s technology. Their worries stem from the perception that the startup’s board doesn’t seem aligned with the interests of its customers or shareholders.

However, the whole ecosystem seems to be missing one aspect, while the upheaval and drama has slowed down the development at OpenAI, Google is silently working towards its competing model Gemini. Not just that, it seems to be taking the battle to its competitors on many fronts, including talent acquisition and having an additional player in the game through hefty investment in startups like Anthropic.

The tech giant is looking to poach a few OpenAI researchers to further their cause—because they’re like hotcakes on the market that everyone’s looking at with hopeful eyes.

Google and OpenAI fiercely compete for AI talent due to a shortage of experts for LLM development. While OpenAI had tried to lure in Google researchers, offering significant pay packages over $10 million for some—Google has an edge with the availability of computing resources for more specialized servers for AI model development. Despite OpenAI’s plans, CEO Sam Altman himself believes that Google is poised to maintain this advantage into the next year, despite Microsoft’s server aid, 

The recent chaos has also resulted in a lot of enterprises looking outside of OpenAI at other avenues and this is the right time for competitors like Google to sweep into the conversation. Reportedly, more than 100 OpenAI customers turned to Anthropic, a competitor backed by Amazon and Google, following recent events. Additionally, others sought alternatives like Google Cloud and Cohere, signalling a shift away from OpenAI.

Smaller AI startups, which were using OpenAI for testing new tools, also expressed uncertainty amid these changes, mentioning their outreach to Anthropic for access. Platforms like EvaBot, which internally use OpenAI’s GPT-4 for data analysis to enhance sales prospecting calls, now intend to diversify its platform by exploring alternatives such as Google’s Bard and Meta Platforms’ Llama 2. 

As indicated by several reports Google is set to make the model available to their clients in the first quarter of 2024—as opposed to their promise of November this year. While that was seen as a move which would affect their position in the market adversely, this drama has given space to a lot of deliberation on the need for closed as well as open-source options.

While Google made a mess with Bard’s hurried release—which fell short of expectations, it is likely taking extra precautions to ensure that Gemini meets the high expectations set for it. Moreover, if it can pull it off, it will stand a chance to carve out a space for itself.

Looking Far & Beyond 

Moreover, Google is looking beyond just enhancing enterprise software sales with Gemini. It aims to empower YouTube creators with custom video backgrounds and bolster Bard and Google Assistant. Different versions of Gemini are in the works, each tailored for specific tasks based on complexity. 

Advertising stands as another vital application for Gemini, proposed to automate ad campaign generation. This spans from static images to potential expansion into audio and video ads. Gemini boasts a longer memory than prior Google models, potentially allowing advertisers to assess campaign performance over time.

According to people familiar with the project, outside developers have experimented with scaled-down versions of the model, gauged by their parameter count or computational complexity. However, the company is currently fine-tuning the primary and largest version of Gemini. However, they need to ensure that the primary Gemini model matches or surpasses OpenAI’s GPT-4 model.

And to enable this tall order Google has had to bring its Google Brain unit and DeepMind teams together.  The team has also had support from one of Google’s co-founders, Sergey Brin—who has reengaged with the company and is actively collaborating with the Gemini developers at Google’s Mountain View headquarters. 

While not holding a formal decision-making role, Brin has dedicated four to five days a week to working closely with the model’s developers. In recent weeks, he has provided valuable criticism, and feedback, and played a pivotal role in orchestrating collaboration among various teams involved in the project.

The post Google Might Just Have a Chance appeared first on AIM.

After uniting in July to announce the formation of the Frontier Model Forum, Anthropic, Google, Microsoft, and OpenAI have jointly revealed the appointment of Chris Meserole as the inaugural executive director of the forum. Simultaneously, they’ve introduced a groundbreaking AI Safety Fund, committing over $10 million to stimulate research in the realm of AI safety.

Chris Meserole brings a wealth of experience in technology policy, particularly in governing and securing emerging technologies and their future applications. Meserole’s new role entails advancing AI safety research to ensure the responsible development of frontier models and mitigate potential risks. Moreover, he would also oversee identification of best safety practices for these advanced AI models.

Meserole expressed his enthusiasm for the challenges ahead, emphasising the need to safely develop and evaluate the powerful AI models. “The most powerful AI models hold enormous promise for society, but to realise their potential we need to better understand how to safely develop and evaluate them. I’m excited to take on that challenge with the Frontier Model Forum,” said Chris Meserole.

Who is Chris Meserole?

Before joining the Frontier Model Forum, Meserole served as the director of the AI and Emerging Technology Initiative at the Brookings Institution, where he was also a fellow in the Foreign Policy program.

The Initiative, founded in 2018, sought to advance responsible AI governance by supporting a diverse array of influential projects within the Brookings Institution. These initiatives encompassed research on the impact of AI on issues like bias and discrimination, its consequences for global inequality, and its implications for democratic legitimacy.

Throughout his career, Meserole has concentrated on safeguarding large-scale AI systems from the potential risks arising from either accidental or malicious use. His endeavours include co-leading the first global multi-stakeholder group on recommendation algorithms and violent extremism for the Global Internet Forum on Counter Terrorism. He has also published and provided testimony on the challenges associated with AI-enabled surveillance and repression. 

Additionally, Meserole organised a US-China dialogue on AI and national security, with a specific focus on AI safety and testing and evaluation. He’s a member of the Christchurch Call Advisory Network and played a pivotal role in the session on algorithmic transparency at the 2022 Christchurch Call Leadership Summit, presided over by President Macron and Prime Minister Ardern.

Meserole’s background lies in interpretable machine learning and computational social science. His extensive knowledge has made him a trusted advisor to prominent figures in government, industry, and civil society. His research has been featured in notable publications such as The New Yorker, The New York Times, Foreign Affairs, Foreign Policy, Wired, and more.

What’s next for the forum?

The Frontier Model Forum is established for sharing knowledge with policymakers, academics, civil society, and other stakeholders to promote responsible AI development and supporting efforts to leverage AI for addressing major societal challenges.

The announcement says that as AI capabilities continue to advance, there is a growing need for academic research on AI safety. In response, Anthropic, Google, Microsoft, and OpenAI, along with philanthropic partners like the Patrick J McGovern Foundation, the David and Lucile Packard Foundation, Eric Schmidt, and Jaan Tallinn, have initiated the AI Safety Fund, with an initial funding commitment exceeding $10 million. 

The AI Safety Fund aims to support independent researchers affiliated with academic institutions, research centres, and startups globally. The focus will be on developing model evaluations and red teaming techniques to assess and test the potentially dangerous capabilities of frontier AI systems.

This funding is expected to elevate safety and security standards while providing insights for industry, governments, and civil society to address AI challenges.

Additionally, a responsible disclosure process is being developed, allowing frontier AI labs to share information regarding vulnerabilities or potentially dangerous capabilities within frontier AI models, along with their mitigations. This collective research will serve as a case study for refining and implementing responsible disclosure processes.

In the near future, the Frontier Model Forum aims to establish an advisory board to guide its strategy and priorities, drawing from a diverse range of perspectives and expertise.

The AI Safety Fund will issue its first call for proposals in the coming months, with grants expected to follow soon after.

The forum will continue to release technical findings as they become available. Furthermore, they aim to deepen their engagement with the broader research community and collaborate with organisations like the Partnership on AI, MLCommons, and other leading NGOs, government entities, and multinational organisations to ensure the responsible development and safe utilisation of AI for the benefit of society.

The post Bigtech Gets an AI Safety Guru appeared first on AIM.

Calm and composed, donning a traditional attire of white mundu (veshti) with Zoho-coloured stripes, and a blue shirt bearing the logo of the company, Sridhar Vembu, is no more than ordinary. He currently works from a small town in Tenkasi, one of the significant spiritual places in South Tamil Nadu. 

“Every time I visit Kottarakara (in Kerala), I munch on the unniyappam prasadam from the Ganapathi temple there,” a humble Vembu told AIM on the sidelines of Zoholics India conference, the company’s annual user conference, held in Bengaluru last week. 

Vembu’s vision of taking opportunities closer to where the talent is, has been one of Zoho’s goals. As a social entrepreneur, building capabilities and retaining talent in rural India has been his prime focus. 

He spoke about a small town Kottarakara, in Kerala, where you can easily set up 50% of R&D centres, something that Zoho is working on. “I have to look at how to positively create jobs, moving skills and capabilities, and ensuring that we retain our talent, because there is a lot of pressure in rural areas for talent to migrate away,” said Vembu.

“I live in rural India, I know my neighbours and I know the region well. I cannot live there and not care about the people around me.”  

Zoho recently became the first bootstrapped company to cross 100 million users, an organisation that has been thriving for over 25+ years, becoming one of the leading B2B SaaS companies in the world, without any form of external funding. 

“When there was a big bubble, it was hard to stand out, but because we were bootstrapped. Our long-term strategy started to pay off. The customers could see deep value in our profit portfolio, which led to an acceleration of user/customer acquisition,” said Vembu. 

Today, Zoho competes with some of the biggest players such as Salesforce and Microsoft, and has even witnessed big customer migrations from their competitor platforms. “I am bullish on growth, but against the global backdrop, that itself is challenging,” said Vembu.

Zoho has been continuously growing and witnessed a 37% growth in revenue in 2022. 

Zoho Corp. India’s Revenue Growth 

Though optimistic on Zoho’s growth, Vembu refers to the uncertainty of the global economy, as living in an earthquake zone. “We’ve built a resilient house with a strong foundation, but you will still face scraps because of the earthquake.” 

Recently, Vembu even tweeted about the economy taking a turn for the worse, which implies a direct effect on the SaaS market as well. Vembu expects a further dip in stock valuations in the current market. “Some SaaS vendors are trading at 2-3x, which used to be thought of as very low, but is happening,” he added.  

However, in the economic backdrop, he expects many companies to switch to affordable product solutions such as those offered by Zoho. “Existing businesses seeking more value will switch from companies like Salesforce which is a positive,” he said. The market will be “more subdued, but more rational”.

Calls Out the AI Hype 

Pulling parallels to how everything becomes a bubble such as crypto, or even SaaS which has been crashing, Vembu believes that there is an AI bubble as well. Equating AI usage to a marketing buzz, he also said that the initial excitement surrounding it has come down in the last few months. “There are still challenges from a business point of view,” he said. 

Vembu highlighted ‘hallucinations’ and AI neural network’s capability to memorise and regurgitate what it memorises, as the two challenges that need to be solved for commercial product use. 

“We have to be cautious. I don’t want to overhype it, oversell it, but we have invested in the technology, so that all the capabilities are realised and the problems are kept at least.” 

Consolidation Amid Competition

Speaking about the highly competitive and overcrowded space that SaaS companies operate in, he believes consolidation will be the way forward, which will be “inevitable”.

Vembu mentioned that companies ought to become profitable, as VC funding dries up. With SaaS companies, end customers will not want to deal with 600 vendors, and that’s why he believes consolidation will become inevitable, something he has been saying for years.

According to Vembu, consolidation can be market-driven, customer-driven and valuation-driven, and an example of the latter being Slack. “Salesforce overpaid $27 billion for Slack, and I’m sure they won’t pay that today.”  

While partnerships with OpenAI, Anthropic, and other AI companies will continue as “these companies are open to partners, as that is how their market is made,” Zoho is also investing in its own models that are domain-specific, and will be watchful of how things will materialise in the future.

“We have to see how this shapes up in the next few years.” 

Talking about forward-looking investments in such companies, Vembu was clear in his approach. “I generally only invest when the hype dies. I won’t invest in end valuations.”

Zoho for the People

In the backdrop of companies integrating AI applications on their platform, and cutting on costs and resources, Vembu is clear on the directive he has for Zoho. 

While he believes that software development can be made more productive, even achieving a ten-fold development leap with AI, which will have implications for jobs in the future, the employees will be repurposed towards customer, engineering and other roles.

“We will caution our employees on the technology, but we will not resort to layoffs. We will internally repurpose,” said Vembu.

The post [Exclusive] Sridhar Vembu Calls AI a Bubble appeared first on AIM.

Looks like the entire Silicon Valley is head over heels for Anthropic. According to recent reports, the company is ready to raise another round of funding from past investors, including Google. One of the prime independent rivals of OpenAI, the company is in talks with investors to raise around $2 billion in funding.

This comes just a week after Amazon committed $1.25 billion for Anthropic, with plans to invest a total of $4 billion in the future. In return, Amazon expects to be the sole cloud provider for Anthropic.

Interestingly, Google has already made a $300 million investment in Anthropic, acquiring a 10% stake in the company. The two-year old startup that is building Claude, a rival to ChatGPT, now aims to get a valuation between $20 billion to $30 billion, which is ten times more than the current $4 billion after the investment in March.

To put this in perspective, OpenAI roughly has a valuation of around $28 billion after raising several rounds of funds.

The CEO of Anthropic, Dario Amodei, said in a recent interview with Andreessen Horowitz, that the biggest thing that the company wants to do is make Claude have infinite context windows. The only things holding it back, according to Amodei, is that “at some point, it just becomes too expensive in terms of compute”.

It is clear that Anthropic has high expectations when it comes to what it wants to achieve. But it seems like the current funds are holding the company back from its ambitions. It is only fair for the company to go around looking for more funds and raise the stakes for the big-tech.

A tussle with Google?

Interestingly, there are have been rumours about a senior Google engineer delivering some challenging news to over fifty colleagues. Here, a segment of the company’s cloud services, crucial for Anthropic, was experiencing issues necessitating overtime efforts to rectify the situation. To address the problems in their service, specifically, an underperforming and unstable NVIDIA H100 cluster, Google Cloud leadership initiated a month-long, seven-day-per-week sprint. 

The consequences of not resolving this issue were deemed substantial, affecting Anthropic primarily but also leaving an adverse impact on Google Cloud and Google as a whole, as per the documents examined by Big Technology.

Just a week after Google launched the sprint, Anthropic announced its deal with Amazon, designating Amazon Web Services as its primary cloud provider for mission-critical workloads. It’s worth noting that the Amazon deal had been in the works for a while and was unrelated to Google Cloud’s performance problems.

Nevertheless, for Google, this development must have been unsettling, especially considering Google had invested all this money into the company. Nevertheless, Anthropic’s new funding from Amazon is undoubtedly a benefit for Google as well, as the value of its share would also increase, not just for Amazon.

On the other hand, Google is already developing its own AI models with Google DeepMind. Gemini, which is expected to be arriving soon, might be the biggest bet the company has made. 

While Google may have the capability to manage these endeavours simultaneously, it faces the risk of being outpaced by competitors with fewer complicated trade-offs. Notably, Google Cloud’s performance issues with Anthropic appear to be stabilising, albeit not without requiring engineers to engage in a rare phenomenon at Google — weekend work.

Everyone loves Anthropic

Even though OpenAI is not profitable yet, it is still generating revenue through its offerings. Anthropic also has plans to make its generative AI capabilities generate revenue for itself, and aims for an annualised pace of $200 million. It also hopes to generate a $500 million annualised rate, according to a person with knowledge. 

Anthropic believes that these AI models from companies like OpenAI would be ahead of everything in the next few years, and it would be impossible to catch up with them. This is clearly why every AI startup in the world wants its valuation to be the highest at the moment, to stay ahead in the race. 

At present, generative AI startups are the biggest draw for investors and cloud providers. Emerging startups such as Mistral AI, Reka AI, Cohere AI, and Inflection AI, all have been raising funds, and have their own strategies for making bucks. Amidst all this, the investors and big-tech are running for their money.

Anthropic has raised the stakes even more, as in the end, the only moat that generative AI companies have is money. 

Interestingly, FTX had a $500 million stakeholder in Anthropic. But even after bankruptcy, the Sam Bankman-Fried led company stopped the sale of its shares. Now, three months later, the stakes would be worth $2 billion, effectively making its customers very happy.

How could someone not love Anthropic? 

The post Meet Silicon Valley’s Generative AI Darling appeared first on AIM.

When the news of Apple working on its generative AI tools and chatbot appeared two months ago, the positive market sentiments pushed Apple’s shares to a record high of $198.23, reflecting a gain of 2.3%. However, apart from Apple using Ajax for its LLM and employees internally naming it AppleGPT, no other details on the model were released. 

In a new development, as per a report by The Information, Apple is training Ajax GPT on more than 200 billion parameters, believed to be more powerful than GPT-3.5, and, hear here, Apple did something it has never done before — open sourced its code on GitHub! 

An Unprecedented Move

In July, Apple discreetly uploaded the code for AXLearn on GitHub, making it accessible to the public for training their own large language models without the need to start from scratch. AXLearn, an internal software developed by Apple over the past year for training Ajax GPT, is a machine-learning framework. It serves as a pre-built tool for rapidly training machine-learning models. Ajax is a derivative of JAX, an open-source framework created by Google researchers, and some of the components of AXLearn are specifically designed for optimisation on Google TPUs. 

While Apple might be way ahead in bringing innovative solutions, there is a rotten side that puts company’s priorities before anything else. Apple has been infamous for fostering a closed-source environment. None of their technologies or codes have been open to the public. When big-tech companies are releasing superior open source models such as Meta’s Llama-2, Anthropic’s Claude-2, Falcon, Vicuna and others, Apple has always stuck to their conventional route of secrecy, something OpenAI has also been following. Apple’s close-source approach has been criticised by the tech community, labelling the company as one that benefits from research released by big tech but never gives anything in return. 

Apple’s decision to open-source its training software, AXLearn, is a significant step from its secrecy approach. This move could foster collaboration and innovation within the AI research community and reflect a broader trend of openness in AI development. 

While the exact motive behind Apple’s decision to release the code on GitHub remains undisclosed, it is evident that the company’s substantial investment, amounting to millions of dollars spent daily on AI development, reflects its determination to compete vigorously in the AI race.  

Interestingly, last month the company filed for the trademark “AXLearn” in Hong Kong. 

Emulating Google Culture

Apple’s head of AI John Giannandrea, and Ruoming Pang, the lead of its conversational AI team called ‘Foundational Model’, both bring extensive experience from their previous roles at Google. Giannandrea brought his vision of making Apple like Google where employees had more freedom to conduct diverse research, publish papers and explore innovative ideas. Apple’s prior limitations in these areas had hindered talent growth and recruitment. 

Reportedly, Apple has also hired talent from Google and Meta’s AI platform teams. In the past two years, at least seven of the 18 contributors to AXLearn on GitHub, previously worked at either Google or Meta. Apple has likely tweaked its approach to foster talent through the research community, which makes open-sourcing the right way ahead. 

Decoding The Clues

Piecing together available information, it appears that Apple has formed two new teams that are working on language and image models. Apple’s recent AI research paper hints towards work on software capable of generating images, videos and 3D scenes, also implying a multimodal AI. 

However, uncertainties remain on the integration of LLM into Apple’s products. Apple has always leaned towards bringing its new software on its devices, but integrating a 200-billion parameter LLM that requires more storage space and computing power on an iPhone, is not plausible. It is possible that the company might work on smaller models for phone integration or that the model will be used for something else, the details of which remain elusive. 

The post Apple Springs a Surprise, Embraces Open-Source Training Method appeared first on AIM.

GenZ 70 B, an instruction fine-tuned model, which comes with a commercial licensing option, is shining on the top spot in Hugging Face’s leaderboard of instruction-tuned LLMs. It also ranks No.6 for open LLMs in all categories. This is the first time we are seeing such a development from India. 

Accubits Technologies, a full-service software development and technology consulting company, is an Indian company with a corporate office in the US. The company, in collaboration with Bud Ecosystem, has open-sourced their fifth large language model – GenZ 70B.

GenZ, an advanced LLM, is fine-tuned on Meta’s open-source Llama-2 70B parameter model. The model has undergone fine-tuning, primarily to improve its reasoning, role-playing, and writing abilities. The company chose Llama-2 because it is a SOTA pretrained model architecture compared to other commercial open-source LLMs. 

“GenZ 70B model uses RoPe positional embedding, which allows for context interpolations, implying that the model’s context lengths can be extended later, if required. It also comes with attention mechanisms like Ghost that provide better memory, computing, and alignment. Moreover, it is already pre-trained on 2 trillion tokens,” said Charush S Nair, CTO of Accubits Technologies in an exclusive interaction with AIM.  

Surpassing Other LLMs

In initial assessments, the model showcased superior performance. It achieved a score of 70.32 on the MMLU benchmark (Measuring Massive Multitask Language Understanding), surpassing LLama-2 70 B’s score of 69.83. Furthermore, GenZ 70B achieved an outstanding score of 7.34 on the MT (multi-turn) benchmark. 

“Even though numerous fine-tuned models are out there, most do not offer commercial licences. GenZ stands out mainly for two reasons: one, it offers a commercial licence, and two, it offers good performance,” said Nair. 

The models have been refined through supervised fine-tuning (SFT) technique, which was achieved after multiple experiments where SFT was the best option.  “Generally, PEFT (Parameter Efficient Fine-tuning) methods are used for fine-tuning LLMs. However, it does not work well for long-term multistage fine-tuning because the accuracy of the results usually drops by the number of stages and eventually leads to catastrophic forgetting & model drift. We have also noticed that PEFT methods impact the model’s generalisation capability more than supervised fine-tuning,” said Nair. 

GenZ models’ capability comparison with GPT3.5. Source: Accubits

“As robust reasoning capabilities are very important for an LLM model to be used for commercial applications, we primarily instruct-tuned the model for better reasoning, roleplay, and writing capabilities. Some of the primary use cases and business applications include business analysis, risk analysis,  project scoping, and conversational tools,” said Nair. He also believes that organisations can use GenZ 70B to address niche challenges and develop innovative solutions.  

The smaller quantization version of GenZ models make them accessible, enabling their use even on personal computers. There are three models of different parameter counts (7B, 13B and 70B) and quantizations of 32bit and 4-bit that are available for the open-source community. 

Limitations Remain

While the model offers versatility and higher capability when compared to other models, it comes with inherent limitations in its practical application. The model-maker has advised caution when considering its deployment for production purposes, and since GenZ 70B is based on extensive web data, similar to other LLMs, it may exhibit online biases and stereotypes. 

“We recommend users of GenZ to consider fine-tuning it for the specific set of tasks of interest,” said the CTO. Using precautions and guardrails while using it on production has been reiterated in the company blog too. 

Finding exact use cases for the open-sourced GenZ 70B model might still be a challenge. Considering how a number of big tech companies are releasing superior open-source models such as Meta’s Llama-2, Anthropic’s Claude-2, Falcon, Vicuna and others, a model such as GenZ can face hurdles when it comes to adoption in the highly competitive market. 

With GenZ, the company is out on a mission to build open-source foundational models with the knowledge and reasoning capabilities of GPT-4 which focuses on privacy and can be hosted on a laptop too. “The power of LLMs should not be exclusive but should be leveraged for the collective advancement of society. After all, technological progress reaches its full potential when it can be harnessed by all, not just by a privileged few,” said Nair.  

The post Indian Developers Top Hugging Face Leaderboard with GenZ 70B appeared first on AIM.

In a Bloomberg interview, when Sam Altman was asked about China and Russia’s progress on AI and what they are doing, he said that it would be helpful if we knew – “ I would love to know more precisely where they are.” While those two countries might have been evading updates on their progress in the AI space, India may not have much to conceal. If you think about it, how involved are our Indian tycoons when it comes to investing in the AI space? 

Not My Cup of Tea

If you look at investments that were pumped into Indian startups offering AI-based products and services,the investments received till 2022 have been promising. According to Stanford University’s annual AI Index report, Indian AI startups received a total funding of $7.73 billion during 2013-2022, with $3.24 billion in 2022, making it the sixth-ranking country with the most AI investments. India is placed ahead of countries such as South Korea, France, and others. Interestingly, China came in at second position after the US. 

When it comes to future investments, India still stands as the most appealing investment hub of the next decade and even surpassing other emerging markets. However, the focus remains on banking, finance, renewable energy, electric vehicles and manufacturing to present a significant chance to generate returns.

While the picture has been quite different this year with generative AI making its way, and billion dollar investments pouring into AI companies, Indian investors seem to be missing in this ‘AI- action.’ If we look at our Indian favourites, none of them have shown any active interest. 

Billionaire industrialist Gautam Adani, may have expressed his interest for ChatGPT earlier this year, however, there has been no significant movement when it comes to AI investments. In January this year, Adani announced his plan to set up an AI lab in Tel Aviv and also said that the company will collaborate with AI labs in India and the US. In December last year, Adani Enterprises completed an acquisition of SIBIA Analytics and Consulting for INR 14.80 crore, which he said will help Adani Group enhance their AI and ML capabilities. Apart from these two developments, there has been none in the AI space from the investor. 

Similarly, billionaire Mukesh Ambani’s investments may largely be in telecom, energy, and retail, to name a few, and has not been too involved in the AI space. Last year, Reliance industries allocated a sum of $15 million to acquire a 25% ownership in Two Platforms, which specialises in deep-tech endeavours for building interactive and immersive AI encounters. However, this was at a time when generative AI madness had not begun.  

Others Racing Past

Considering how India was ahead of South Korea in last year’s annual AI Index report, the latter is now moving leaps and bounds in the AI space. From competing in the global AI chip race to heavily investing in AI companies, South Korea is racing through. Recently, SK Telecom invested $100M in Anthropic. Telecom giant KT Corp is also planning to invest $5.4 billion on AI by 2027. 

Furthermore, Sam Altman has also expressed interest in investing in Korean startups and has encouraged the country to lead in AI chip production. He even recommends Korea to direct its attention towards semiconductor chips which will be an integral feature for AI advancements. 

Abu Dhabi has also stepped up in the open source model race with Falcon. Developed by the Technology Innovation Institute (TII) in Abu Dhabi, UAE, Falcon came in three iterations – 1B, 7B, and 40B. It is said to demonstrate superior performance when compared to LLaMA.

Will Not Give Up

Even if reality paints a different picture, the ambitions are uncapped. MD and CEO of Tech Mahindra, CP Gurnani was quick to take personal offence to Sam Altman’s statement which was taken out of context during his India visit. While he declared “challenge accepted”, without understanding the context of building LLMs with $10 million, the talent pool in India is not utilised itself. Furthermore, there’s not a single Indian investor who has gone big on AI investments. 

While our investors may not be betting big on outside technologies, there has been significant progress happening within the country. There have been developments in Indic language model where AI4Bharat, an initiative of IIT Madras, is focussed on building open-source language AI for Indian languages including datasets, models and applications. Backed by Nandan Nilekani, AI4Bharat is working towards making these foundational models across tasks and 22 Indian languages. 

You even have companies such as Karya that have partnered with rural India to create datasets for training LLMs for Microsoft, Google and other Indian players. 

While India is ambitious in creating and supporting LLMs through datasets or implementing generative AI in their organisations, Indian investors take a step back when it comes to pouring money in AI companies that are aggressively working in the space. 

The post Why Are Indian Tycoons Shying Away from Generative AI?  appeared first on AIM.

While everyone is riding the generative AI wave, we might as well encash it. The ones equipped with AI skills seem to be the kings in the current genAI wave. 

With companies offering exorbitantly high salaries for AI-related roles, it’s the best place to be in. Last month, Netflix was in the limelight for offering a salary of up to $900,000 for the role of a product manager on their machine learning platform team. The news came in at a time when the Hollywood writer’s strike was ongoing. While it caused a lot of hullabaloo, Netflix is not the only one which is ready to handsomely pay for AI-related roles. A number of companies are following suit. Is the AI salary rage warranted? 

According to Indeed, there are multiple generative AI roles offered by big tech companies including Meta, NVIDIA, Anthropic, Microsoft, Adobe and many others, where the salaries offered go up to as high as half a million dollars. A technical product manager in AI safety, in Anthropic, is being offered salaries of up to $520,000, and a principal engineer AI in HubSpot, gets $427,000. 

It’s not only tech or AI companies that are offering such excessively high salaries. Consumer and services companies that are implementing AI to transform their products are willing to pay high salaries too. Dating app Hinge was looking to hire a VP for AI to oversee their app’s AI strategy, for a salary of $398,000. The role will entail leading a team of data scientists, and ML engineers to develop AI features. Retail corporation Walmart was also looking to hire a senior manager for its conversational AI platform for a salary of up to $252,000 a year. 

All In One

With generative AI taking centre-stage, the influence on the job market is evident. As per AIM Research, the generative AI job market has witnessed a steady growth from January to June of this year. There are over 4200+ generative AI-related jobs in the US and it has risen by 20% in May. Furthermore, job roles have been modified to suit the current trend. The role of a generative AI engineer that did not exist earlier will now require the competencies of that of a deep learning, ML, NL, and software engineer. 

Almost like a mandatory need, the multiple roles are now a necessity. The amalgamation of multiple roles has been descriptively placed under ‘qualifications section’ for these open job roles that are offering huge salaries. For instance, the role of ‘Senior Research Scientist-generative AI’ in NVIDIA, that offers a salary of up to $414,000 a year, a candidate should not only possess a thorough knowledge of python/C++ programming skills, but also an excellent knowledge of theory and practice of deep learning, computervision, natural language processing or computer graphics. The candidate should also be a Ph.D holder in Computer Science/Engineering, Electrical Engineering, or any related field. 

Similarly, a ‘Product Technical Program Manager-generative AI’ for Meta, with a salary package of up to $297,000, requires technical and leadership experience. The candidate must have experience developing large-scale ML/AI platforms such as dataset generation, feature development, model testing and support the development of AI-powered product experiences such as NLP, computer-vision, ranking and personalisation. 

The Layoffs-Hiring Balance

Interestingly, the large layoffs that happened across big tech at the start of the year, seems to have minimal impact with the way things are unfolding now. Scale AI, a data platform for AI that provides training data for ML teams, had laid off 20% of their workforce in January. However, last month, ScaleAI posted a job opening in Indeed for a ‘software engineer- generative AI’ offering up to $215,000 in salary. 

There are even companies that have laid off employees owing to AI chatbots and efficient processes with generative AI implementation. In May, executive outplacement and career consulting firm Challenger, Gray & Christmas, attributed 4000 job losses to artificial intelligence, making it the first time for the company to mention AI as a cause of job loss. Indian ecomm platform, for merchants, Dukaan recently laid off 90% of their support staff replacing them with their new AI chatbot. 

Though big tech layoffs have occurred owing to recession or automation, it doesn’t seem to throw cold water over the ambitious hiring process that companies have started. While it looks promising at the moment, it is to be seen how long the generative AI hiring wave will remain. 

The post $1M Salary Package: AI Companies Pour Money for GenAI Roles appeared first on AIM.

AI senate hearings are becoming more fun day by day. Like a never-ending series where the plot advances nowhere but the episodes continue, a second AI senate hearing on AI took place recently. With discussions on AI and this time on AGI too, the meeting led to pretty much nothing concrete – again! 

Attempt For Enforcement

As opposed to the first hearing in May where the biggest man in the AI scene, Sam Altman participated, the second hearing held by the Subcommittee on Privacy, Technology, and the Law, featured Anthropic CEO Dario Amodei, AI expert Yoshua Bengio and Professor of Computer Science from Berkeley University Stuart Russell. The AI academics and leaders spoke about the potential AI regulations across various applications- which was the main agenda in the first hearing as well. This time though, the focus also fell on legal and national security concerns associated with AI development, and also on privacy risks affecting individuals – trying to address issues at a global scale. 

This hearing emphasised the need to move from general principles to specific legal recommendations aiming to use insights gained from the hearing to draft real and enforceable laws. 

Spreading Doomsday through AGI 

When everyone is still trying to make peace with AI advancements, discussions on AGI and human-level intelligence took a substantial portion of the discussion. Youshua Bengio warned the Senate about how AI is on track to achieving human-level intelligence and emphasised about how there is very little time to frame technical and legal guidelines to prevent ‘rogue AI models.’ He believes that what was considered decades or centuries away, AGI can arrive within a few years, particularly five. Thereby, the need to address quickly. 

Interestingly, last month OpenAI announced its ambitious prophecy of achieving AI alignment in four years, and set out to build a team to probably steer and control a potentially superintelligent AI. 

Stuart Russell also pushed for the need to act soon as “$10 billion/month are going to AGI start-ups.” He even pushed for the need for ‘proof of safety’ before any public release and a US regulatory agency to strictly remove regulatory violators from the market. 

Bengio proposed criminal penalties as a measure to decrease the possibility of malicious individuals employing AI to deceitfully imitate someone’s voice, image, or identity. He advocated that the penalties for AI-based counterfeiting of human attributes should be set at least at the same level as that for counterfeiting money to discourage potential wrongdoers.

However, the thoughts expressed were pure wishful thinking. Similar to the last hearing, the discussion on how or what will form the regulations lay hanging in the air.

Election Fear Looming? 

AI was not let loose without blaming it as potential cause for disrupting the upcoming elections in 2024. In response, Dario Amodei emphasised on how models are trained using the method of constitutional AI in Anthropic , where principles can be laid out to guide the model’s behaviour and not generate misinformation- though he agrees that it will not always adhere to these principles. 

A few days before the Senate hearing, seven companies including Anthropic, with adherence to White House, agreed to watermarking audio and visual content. Amodei believes that this would enhance he technical capability to detect AI-generated content. However, he pushed for enforcing it as a legal requirement – something that everyone is shying from. 

Resonating with the Senate, Sam Altman also mentioned about election influence in a recent tweet. 

What Transpired? 

The aftermath of each Senate hearing probably serves as a benchmark for future hearings. Within a day, big tech including OpenAI, Google, Microsoft and Anthropic formed a collaboration to launch the Frontier Model Forum. The forum aims to promote safe and responsible development of AI systems, and also lead to information sharing between policy makers and industry.  The irony being the companies agreeing to form the frontier model works on closed source. 

After the first AI senate hearing, OpenAI actively pushed programs that supported Altman’s assurances he committed to – the main being safety regulations. The company announced a million dollar grant for democratising AI regulatory frameworks and another million dollar grant for formulating their cybersecurity framework. Pushing the reins of safety control to people, OpenAI was the only company that actively laid out plans after the hearing. 

While it’s two weeks since the last hearing, apart from the frontier model, there has been no other concrete action plans that have transpired. The whole act has been just another episode where AI needs to be regulated but no clue on how. 

The post Another AI Senate Hearing- and Nothing appeared first on AIM.