AI Update: Robots, National Security, and Cross-model QA

It may be the holiday season, but artificial intelligence never sleeps. Here’s the latest big news on the shifting landscape of AI and identity technology:

xAI has completed a $6 billion Series C funding round, which featured A16Z, Blackrock, and Sequoia Capital, among others. Nvidia and AMD also participated as strategic investors. The funding round is thought to value Elon Musk’s AI startup in the range of $35 billion to $40 billion.

The Dutch government plans to include AI in an investment screening law, meaning that foreign entities wishing to acquire AI-related firms, tech, or infrastructure will be under additional scrutiny, with a review process that could take six to eight months. The government views it as a national security measure.

OpenAI’s leadership has reportedly been discussing the development of humanoid robots. The firm has rebooted an internal robotics software team that had been shut down four years earlier, and has recently been investing in robotics startups including Figure and Physical Intelligence.

Palantir and Anduril are in talks with OpenAI and others to form a consortium that will bid on Pentagon contracts. The news comes after the announcement of a partnership between OpenAI and Anduril that will integrate the former’s AI technology into the latter’s weapons systems. Other potential partners include Elon Musk’s SpaceX, Scale AI, and Saronic.

Broadcom CEO Hock Tan believes the AI spending frenzy will continue to the end of the decade. He expects Broadcom customers to build clusters with up to 1 million AI chips by 2027 to help keep up with scaling needs, and he thinks his firm’s AI chip sales could reach tens of billions of dollars within that same timeframe.

Google’s contractors are comparing its Gemini AI output to that of rival Claude from Anthropic in their efforts to assess quality. They’re not just looking at accuracy in responses, but also adherence to safety guidelines, noting in internal correspondence that “Claude’s safety settings are the strictest”. Google recently unveiled a much-improved version of Gemini, as well as a powerful text-to-video model.

The chatbot’s take: Rather than expose ourselves to the risk of being wrong, we asked GPT-4o to make some predictions about what’s in store in 2025:

–

December 27, 2024 – by Alex Perala