☕️ Anthropic's Claude 3.5 Sonnet Blows GPT-4o Out of The Water

Anthropic keeps its competitors on their toes as it releases Claude 3.5 Sonnet, a massively improved version of its middle-of-the-pack Claude 3 model, keeping 3.5 Opus and Haiku as potential trump cards. Meanwhile, Ilya Sutskever remains ever-vigilant in pursuit of safe ASI by launching a company dedicated to it, Nvidia becomes the world's most valuable company, surpassing Microsoft, and UK and German researchers leverage AI to diagnose Parkinson's early.

Join us at AI Tangle as we untangle this week's happenings in AI!

Anthropic comes out swinging with Claude 3.5 Sonnet, decimating benchmarks

Anthropic has launched its latest AI model, Claude 3.5 Sonnet, seemingly out of the blue to top the likes of OpenAI's GPT-4o and Google's Gemini. This version of the model is available to users on the web and iOS, as well as making it available to developers. Claude 3.5 Sonnet is positioned as the middle-of-the-pack model in Anthropic's Claude 3 lineup, outperforming its predecessor, Claude 3 Opus, by a significant margin - but how impressive is it really?

​

What did Anthropic achieve with Claude 3.5?

Claude 3.5 Sonnet excels in various benchmarks, outperforming GPT-4o, Gemini 1.5 Pro, and Meta's Llama 3 400B in seven of nine overall benchmarks and four out of five vision benchmarks. It is designed to improve tasks such as writing and translating code, handling multistep workflows, interpreting charts and graphs, and transcribing text from images. Claude 3.5 Sonnet's improvements over Claude 3 Opus are especially noticeable in math and coding benchmarks. Additionally, Anthropic introduced a new feature called Artifacts, allowing users to interact with and edit the results of their Claude requests directly within the app.

Former OpenAI chief scientist Ilya Sutskever launches an AI company to develop a safe superintelligence

Ilya Sutskever, OpenAI's former chief scientist and a former co-lead of its Superalignment team, recently ​debuted his latest venture in an X/Twitter post​ named ​Safe Superintelligence​ (SSI for short). SSI was co-founded by Y Combinator partner Daniel Gross and ex-OpenAI engineer Daniel Levy with the sole goal of developing a safe superintelligence. Currently, the startup is recruiting technical talent at Palo Alto and Tel Aviv, and though SSI hasn't addressed any funding or valuation problems, Levy says that "raising capital is not going to be one of them."

​

Nvidia overtakes Microsoft to become the world's most valuable company at $3.33 trillion

Just a few weeks after ​overtaking Apple​ to claim the spot of second most valuable company in the world, AI kingpin Nvidia has done it again and overtaken Microsoft for the throne. As of writing, the company's shares are priced at roughly $136 following the 10-to-1 split on the 7th of June, putting the company's market cap at $3.33 trillion. Nvidia's H100 GPUs have been dominant in training AI models internationally, and the company's ​latest Blackwell B200​ might widen the gap between it and competitors even more.

​

UK and German researchers developed an AI test to diagnose Parkinson's early

Researchers from University College London and University Medical Center Goettingen have developed a simple blood test that uses AI to predict Parkinson's disease up to seven years before symptoms appear. The test identifies eight protein markers linked to inflammation and protein degradation, allowing for early diagnosis and potential treatments. Future plans include creating an even simpler test, where a drop of blood on a card can be posted to the lab, to see if it can predict Parkinson's even earlier.

​

Apple faces numerous challenges for its Apple Intelligence launch in China

Apple's recent Apple Intelligence system made waves at its WWDC 2024 keynote, but the company faces challenges deploying it in China. During the WWDC 2024 keynote, Apple said that its on-device Apple Intelligence would roll out in the autumn of this year in US English and other languages, but mentions of China were nowhere to be heard. China's AI and privacy regulations are stringent, and Apple choosing to partner with OpenAI to leverage ChatGPT, which is banned in China, means that a China-specific Apple Intelligence would most likely require a domestic partner.

​

SoftBank plans to develop an AI-powered system to "cancel out" angry customers' emotions during voice calls

Customer support workers often have it tough with angry users, but SoftBank recently announced that it is developing an AI-powered system to "cancel" such emotions during phone calls by altering their tone and pitch. The goal of this is rather obvious: to ease the psychological burden on customer service representatives. However, some social media users claim this could be a sign of a bigger problem if a call center is receiving so many angry complaints to begin with and that "ignoring reality" doesn't address the root cause.

​

UK and France dominate the European market in GenAI startups and funding

Generative AI in Europe and Israel is experiencing a recent surge, though it is still lagging behind the US in funding. A study by Dealroom and Accel found that London boasts the most startups, while France surprisingly dominates funding with a staggering $2.29 billion, ahead of even Israel. The Frenchman's success is likely due to a strong talent pool nurtured by top universities and the presence of tech bigwigs like Google and Meta building up research labs, creating a fertile ground for future founders.

Jellypod

Too many newsletters? Listen to them on your commute. Jellypod AI (iOS) creates a personal podcast every morning of all of your favorite newsletters. Download, select your fav newsletters, and done.

​Airtable - Categorize information, generate drafts, and translate content with Airtable AI, transforming your operations and embedding AI into everyday work.

​Retention - Up your language level skills and maintain existing ones with Retention, an AI-powered learning system for intermediate and advanced language learners.

​Olvy - Surveys, interviews, reviews, support tickets and sales calls - bring them together into one workspace for faster insights using state-of-the-art automated AI by Olvy.

​UPDF AI - UPDF AI streamlines PDF interactions and research by leveraging AI and offering detailed summaries, translation, and Q&A chat with your documents to boost the work you can get done.

​Jensen Huang's Vision of an AI-Powered Electric Grid (3-min read)

In a talk at the Edison Electric Institute, Jensen Huang, founder of Nvidia, spoke about ways the new industrial revolution of generative AI is transforming the future for utilities and their customers.

Ilya Sutskever & The Future of SSI (12-min watch)

Wes Roth, a prominent YouTube channel covering the latest in AI developments, dissects the latest news of Ilya Sutskever and his latest venture, Safe Superintelligence (SSI), in pursuing, you guessed it - safe superintelligence.