AI Update: Jagged AGI and Tiny Worm Brains

Welcome to the newest edition of ID Tech’s AI update. Here’s the latest big news on the shifting landscape of AI and identity technology:

OpenAI unveiled its newest, highly intelligent AI model, o3, at the conclusion of its 12 Days of OpenAI event. Prominent commenters on X are, at the time of writing, disagreeing somewhat about just how highly intelligent it is, with some exclaiming that “AGI has been achieved” while others are suggesting something like “Jagged AGI – superhuman at some tasks, weaker at others.” Coders see a model that is coming for their jobs:

Liquid AI has raised $250 million in a Series A funding round. The MIT startup was founded just last year, and is working on AI systems that are modelled not on the human brain, but on that of a tiny worm called Caenorhabditis elegans, which is only 1 millimetre long. The funding values the startup at $2.3 billion.

US-based Vultr has raised $333 million in its first financing round, led by LuminArx Capital Management and AMD Ventures. The firm was founded in 2014 and has so far been self-funded. It describes itself as the “leading alternative hyperscaler” and the “world’s largest privately held cloud computing platform.” It makes $150 million in annual recurring revenue, and is now valued at $3.5 billion.

The US Department of Commerce has asked Nvidia to investigate how its AI chips ended up in China despite sanctions. Nvidia has, in turn, asked its major distributors like Dell and Super Micro Computer (Supermicro) to do spot checks of their customers in Southeast Asia.

Microsoft has acquired about twice as many Nvidia chips as its rivals over the past year, according to an analysis from Omdia. The stockpile includes about 485,000 of Nvidia’s “Hopper” chips. Meta is in second place, with 224,000 Hoppers, followed by Amazon (196,000 chips) and Google (169,000 chips).

Lockheed Martin has launched a subsidiary that will focus on helping US defense contractors integrate AI into their operations. Astris AI‘s announcement comes as the incoming Trump administration prepares to take advice from a Department of Government Efficiency co-chaired by Elon Musk, and shortly after OpenAI entered into a partnership with Anduril, a Silicon Valley defense technology startup.

A newly released audit has found that evolving regulatory frameworks are slowing down the FBI’s ability to adopt AI technology, with a growing backlog of AI use cases waiting to be approved. The audit also found that the FBI has been encountering a lack of transparency on the part of AI vendors, especially those offering solutions with embedded facial recognition capabilities.

Ukrainian military authorities are pitching a trove of video data from their drone strikes for AI training, with the aim of developing more sophisticated AI weapons systems. They say they have now collected over 2 million hours’ worth of footage. Speaking to Reuters, an official with the U.S.-based Center for a New American Security said it could offer valuable data for combat against Russia, but noted that US officials are prioritizing data that can help in the Pacific context.

The chatbot’s take: For what it’s worth, the o1 model isn’t calling it AGI. But is o1 just being defensive?

–

December 20, 2024 – by Alex Perala