Hey there, folks, welcome to the latest AI newsletter from TechCrunch. If you want to receive this in your inbox every Wednesday, sign up here.
The battle of the billionaires is back.
Elon Musk, the world’s richest man, made an offer to buy the nonprofit governing OpenAI for a whopping $97.4 billion. OpenAI CEO Sam Altman responded with a witty post on X, saying, “No thanks, but we’ll take Twitter for $9.74 billion.” Musk and investors bought Twitter for $44 billion in 2022.
Whether Musk’s bid is serious or not, it could complicate OpenAI’s move to become a for-profit public benefit corporation within two years. The board will have to ensure they’re not selling the nonprofit short by handing its assets, including research IP, to an insider like Altman at a discount.
OpenAI could argue that Musk’s offer is a hostile takeover since he and Altman aren’t on good terms. It could also question whether Musk has the funds. In response, OpenAI’s board stated that the bid doesn’t set a value for the nonprofit and that it’s not for sale.
In a statement Tuesday, Andy Nussbaum, representing OpenAI’s board, said Musk’s bid doesn’t determine what’s best for OpenAI’s mission.
My colleague Maxwell Zeff and I wrote a detailed piece on what to expect in the upcoming weeks, but rest assured, Musk’s offer and the ongoing lawsuit against OpenAI promise some intense legal battles.
News
Image Credits: Apple
Apple unveiled a research robot that works like a more dynamic HomePod. Users ask questions to the lamp, and it responds in Siri’s voice.
AI for All?: An essay by Altman discussed how AI benefits aren’t evenly distributed and mentioned OpenAI’s openness to innovative ideas like a “compute budget” for broader AI usage.
Artificial Controversy: Christie’s plans to host a show dedicated to AI-generated art, sparking mixed reviews and a petition for cancellation.
Google’s Golden AI: Google DeepMind’s AI system surpassed gold medalists in solving geometry problems in a math competition.
Research Paper of the Week
Image Credits: MIT CSAIL
A research team at MIT CSAIL found that over 50% of errors in top AI models stem from mislabeled and unclear questions in popular benchmarks. They suggest rethinking benchmarks to ensure model reliability.
Model of the Week
Image Credits: kudzueye
Boreal-HL is an AI video generator specializing in mundane scenes like tourists eating ice cream or people at lunch meetings with a funny twist.
Innovations in AI efficiency are making highly sophisticated models cheaper and easier to train. A study by researchers at Shanghai Jiao Tong University and SII showed that a model trained on 817 samples outperformed those trained on 100 times more data.
Stay tuned for more exciting developments in the world of AI.
