AI & Startups October 31 - November 5

Gen AI Product News

Cohere debuts Embed v3 for search

AI startup Cohere released a new model called Embed v3, a major industry upgrade for querying noisy real-world data. The release also significantly boosts retrieval for AI writing tools, while the model’s multilingual version supports 100+ languages for searching across dialects.

Apple bets on AI for next-gen health monitoring

Apple reportedly plans to launch advanced health monitoring powered by AI for its Apple Watch and AirPods, as well as paid health coaching services in 2024.

The details:

• Apple Watch will gain sensors and algorithms to detect conditions like hypertension and sleep apnea without dedicated devices.
• Apple aims to leverage AI speech analysis and audio tuning to enable AirPods to serve as hearing aids and provide hearing tests.
• A new subscription service will offer personalized workout and nutrition recs, powered by users' health data and AI models.

The Beatles' release final song revived with AI

Paul McCartney and Ringo Starr just released a new Beatles single featuring AI-generated John Lennon, completed using machine learning to isolate individual tracks from an unreleased demo.

The details:

• "Now and Then" stalled in the 1990s because analog tape technology couldn't separate Lennon's blended piano and vocals.
• New AI audio processing extracted Lennon's parts to allow McCartney and Starr to overdub new bass, drums, guitar, and strings.
• The result features all four members' performances, with McCartney calling it ‘likely’ the final new collaboration featuring the full band.

LinkedIn 's AI job hunter arrives

LinkedIn began rolling out an AI-powered "job seeker coach" for premium members, alongside an announcement of surpassing one billion users on the site. The tool aims to match job candidates with openings, prep for interviews, and more —?and comes as AI and a potential recession raise hiring anxiety.

GPT-V API access coming soon?

If you’re hoping for API access to ChatGPT’s latest voice and image capabilities, you may not have to wait much longer. OpenAI Head of Dev Relations Logan Kilpatrick tweeted “soon”, fueling speculation of an announcement at the company’s Dev Day on Nov. 6.

Google's MetNet-3 AI improves weather forecasts

Google just released MetNet-3, an AI model that makes highly accurate 24-hour weather forecasts. Google claims MetaNet-3 outperforms the top traditional physics-based models, with the upgrade now live across Google’s weather-related products.

Google DeepMind 's AlphaFold update advances drug discovery

DeepMind just revealed an update to AlphaFold 2, which can now predict structures of diverse molecules beyond proteins — unlocking new drug discovery capabilities.

Key points:

• AlphaFold can now predict 3D structures for nearly all cataloged molecules in biology — including DNA, RNA, and small molecules.
• The update unlocks a new understanding of complex systems like protein interactions, cell signaling pathways, and genome editing tools.
• Researchers anticipate AlphaFold’s upgrade has the potential to rapidly advance disease research, drug development, synthetic biology, and other scientific fields.

Poe launches AI chatbot creator monetization

Quora 's AI chatbot platform Poe just introduced a revenue-sharing program to pay bot creators when their bots lead to app subscriptions.

Key points:

• Bot creators earn up to $20 per monthly subscription or $10 per annual subscription resulting from a user engaging with their bot.
• Creators can soon set per-message fees that Poe will pay out for each user interaction, providing recurring revenue.
• The program aims to incentivize developers and researchers to build differentiated bots that drive new users to Poe.
• The move follows the rise of lucrative creator economies on social platforms — and could expand to other AI assistant apps.

AI company Phind claims that its model can beat OpenAI’s GPT-4 at coding

After Chinese tech giant Baidu recently claimed that its AI chatbot Earnie 4.0 can match ChatGPT for performance, Phind now claims that their new model can beat OpenAI’s GPT-4 at coding.

Phind claims that their new model “matches and exceeds GPT-4's coding abilities while running 5x faster” — a huge claim. However, the company admitted that one area where the model struggles is consistency, and said that it requires more generations to find the correct answer compared to GPT-4.

Code assistance has been one of the leading use cases for generative AI, with Github’s Copilot becoming one of the first generative AI products to exceed $100 million in revenue.

You can find more technical details about Phind’s new model

xAI’s ‘Grok’ chatbot will be available to X Premium+ subscribers only

Leaker claims Instagram is working on a customizable AI friend feature

Making friends with an AI seems a little out there for most people. But AI chatbot apps like Character AI are now generating hundreds of millions of monthly visits from eager users who spend hours every day talking to AI-generated characters.

And after the launch of celebrity AI chatbots across its apps last month, Meta might be working on a new feature that lets users create their own AI friends, according to leaker Alessandro Paluzzi on X.

According to the screenshots shared by Paluzzi, users will be able to pick the ethnicity, gender, age, personality and interests of their chatbot friend

Gen AI VC

Defense Tech Startup Shield AI Raises $200M At $2.7B Valuation

Nvidia Co-Leads $99M Round For 3D Metal-Printing Startup Seurat Technologies?

German AI, deep tech startup TVARIT bags €6.5M to expand zero waste metal manufacturing

Dashtoon raises $5M to deploy generative AI for comic creation

AI Regulation

Nations come together on AI safety at historic UK summit

The United States, the EU, China, and over 25 countries signed a declaration to collectively govern AI risks on day 1 of U.K. Prime Minister Rushi Sunak’s AI safety summit, marking a milestone in oversight.

The details:

• China was included in this first major West-led effort on AI safety, agreeing to increased collaboration.
• The ‘Bletchley Declaration’ lays the groundwork for joint risk assessment and policies to ensure safe, ethical AI development.
• Critics question if geopolitical rivals can truly cooperate on AI — but the initial willingness signals some common ground.
• Elon Musk, who attended the AI Safety Summit, stated he believes AI is 80% likely beneficial and 20% dangerous.

Biden issues U.S.′ first AI executive order, requiring safety assessments, civil rights guidance, research on labor market impact

AI Events

OpenAI Dev day keynote livestream on Monday at 10AM PST.