Weekend Warp
Jaspreet Bindra
Founder: AI&Beyond and Tech Whisperer Ltd | ex-CDO Mahindra, Microsoft, TAS | Author - The Tech Whisperer | Faculty - AshokaU, SingularityU | M St - AI & Ethics, Cambridge University | Gurgaon, Cambridge, Dubai
The last week belonged to Ilya.
Ilya Sutskever , OpenAI cofounder, protégé of Geoff Hinton, and the OG AI legend, broke away from his friend Sam Altman to set up his own outfit Safe Super Intelligence or SSI. ?SSI, as the name suggests, promises to have one goal and one product – Safe Super Intelligence. “We plan to advance capabilities as fast as possible while making sure our safety always remains ahead. This way, we can scale in peace.” says it’s manifesto. Last week SSI and Ilya got a massively scaled shot in its arm, when they raised a $1bn initial round at? $5bn valuation from Andreessen Horowitz , Sequoia Capital , DST Global and others. 3 months in, 10 employees, not even a website, and a billion dollars in the bank. That is the power of the idea and, more than that, the name of Ilya Sutskever.
Altman, meanwhile, is not holding back. He teased an X post where he was growing strawberries in his garden, hinting at a powerful new project coming out. Initially called Qstar, Project Strawberry seeks to inculcated reasoning and some mathematical abilities into GPT4o, and give it what CTO Mira Murati calls, “PhD level intelligence”. Today’s LLM can do math but not the traditional way. They cannot actually add or reason 2+ 2 to give 4, they would give the right answer most of the time since the vast WWW based databases they are trained on probabilistically would give 4 as the next word, and therefore the answer, ?when you say 2 + 2 = ? This, obviously, breaks down with larger more complicated mathematics and reasoning. This has been the ultimate challenge for LLMs to become ‘intelligent’, whether safe or not, but Strawberry is rumored to solve for that. Interestingly, it is further rumored that it was Qstar which prompted Ilya to participate in the boardroom coup against Altman and then finally leave, resulting in the oft repeated meme, ”What did Ilya see?”
领英推荐
Moving further down Palo Alto to Mountain View and further across to London, where Demis Hassabis reigns over Google DeepMind and all of Google AI, came another announcement. Continuing the Alpha series of products, Alpha Proteo was launched as an AI Model to generate new proteins using AI. Google DeepMind recently launched AlphaProteo, an AI system that generates novel proteins designed to bind to specific target molecules poised to significantly advance research in drug design, disease understanding and other health applications. This AI system aims to create new protein binders for diverse target proteins, including those linked to critical health issues like cancer. While AlphaFold, its earlier product could predict how proteins fold, this one can actually generate new ones. I have always maintained that it is in this area that the real miracles or AI are being seen, not necessarily the buzzy chatbots.
The space is certainly heating up....... at warp speed.