Weekend Warp

After a mega week the week before, AI stopped for a breath this week. Relatively speaking, that is. The usual characters – Altman, Musk, et al – were in the news, though Microsoft and Nvidia took a break, handing the baton to Apple for a change.

So, let’s start with 苹果 . They seem to have bit on the forbidden fruit by inking a potentially huge deal with OpenAI to have a GPT-powered Siri. Forbidden, because it is an admission that they cannot build on their own, and also because it means indirectly allying with old foe 微软 , which is OpenAI’s main backer. For Sam Altman, personally, and for OpenAI this is a huge, huge deal. Much like 谷歌 Search has been default on the billions of iPhones sold, the same would happen with ChatGPT. I had written recently about how Google will try to out-distribute OpenAI, rather than out-innovate it; with this deal, OpenAI gets a humongous distribution base of 1.5bn people, about 30% of global smartphone users. It will be interesting to know who will pay whom, though, and how Apple will deal with hallucination and security issues. Also interesting is Microsoft’s reaction – Satya Nadella is reportedly quite nervous about this deal as it means its strongest partner is allying with their arch enemies, as also whether the cloud infrastructure can handle this sudden additional load. The deal is to be announced in WWDC.

Read more here

Meanwhile, storm clouds of a different kind seem to be gathering over OpenAI and Altman. After co-founder Ilya Sutskever and Jan Leike, who handled the Safety and Alignment teams at OpenAI left with the latter openly rubbishing OpenAI’s AI safety efforts, its ex-Board member Helen Toner finally spoke up on why the Board voted Altman out. She revealed that the Board came to know about the ChatGPT launch on Twitter , and about Altman’s refusal to share information, his tendency to move away from the original purpose and structure of the company, governance issues, and the pressure to build larger models for profit, no matter what the consequences would be. Read more here .

I have written about this in Mint here, where I say the following: “ The problem is usually never the technology but the business model. Data-monetization as a business model has sunk social networks into the morass they are in, with X’s owner Elon Musk railing against advertisers who seem to be “blackmailing" him and his noble intentions for X. Google succumbed to ‘the innovator’s dilemma’ as it was afraid that the large language models of AI it had worked on could hurt its lucrative search business model. So will it be for Generative AI and OpenAI. The company twisted itself into a pretzel-like corporate structure, enabling it to raise billions of dollars from Microsoft and attain a valuation placed at $86 billion, while simultaneously claiming to build AGI for the good of humanity and not necessarily for economic gain. These divergent forces could not be sustained forever and the pretzel model broke down when the strain grew too severe. The outcome was clear: capitalistic business models had won over the quest for pure societal good.”

Now that the Altman quota is over, let’s move to Musk. Fresh from wrapping up a $6bn round for X.ai , he has unveiled plan to create an AI Giga Computer, also called the ‘Gigafactory of Compute’ by 2025. It will have 15000 英伟达 chips, and he plans to partner with 甲骨文 for that (interestingly because the three big hyper scalars are competitors?), ?it will power the next version of his quirky chatbot, Grok. The space gets more and more interesting

Read more here

Until next week, then.