Weekly Links: Jensen Huang, AI Overwhelm, and the Roblox Metaverse
This week saw the resignation of Jensen Huang, air traffic control outages in the US (though possibly not AI related), and AI self improvement acceleration.
This week in AI: NVIDIA CEO Jensen Huang resigned (see below), the FAA grounded all flights in the US for an extended period, and an AI company had to announce it was not to blame. Anthropic also announced it reached its "security threshold" for AI self-improvement, but it's entirely unclear what will change as a result.
This week is a little shorter because of personal commitments this weekend. Just three main stories:
- Nvidia CEO Jensen Huang resigns in blow to AI boom he helped foster. Jensen Huang's resignation this week is hard to process given how ever-present he has been in the AI Boom. NVIDIA plays an instrumental role not only in the chip layer of AI but throughout the ecosystem through funding, acquisitions, and a great deal of support for open source, which is in its interests. NVIDIA has famously been described as an F1 car built around Jensen Huang; it will be hard for anyone to sit in the driver's seat.
- An Interview with Roblox CEO David Baszucki on Aggregating Play. There is a lot of smart stuff here (where they see the tech going - basically remote multiplayer as meshes with local photorealistic overlay - their strategy to pull in other games and a move into allowing single-player). Roblox is definitely the dark horse for the metaverse.
- ‘AI clog’: UK public bodies battle onslaught of chatbot-generated complaints. This might seem like a minor concern: increasing inflow to public services, grant bodies and other institutions from AI-generated applications. However, it seems likely that all this inflow will continue to explode, and we haven't even gotten to the main part of the growth curve yet, when voice agents are used to call phone lines (this is already happening, but not at scale). Clearly, users are asserting their rights, whereas friction prevented them from doing so before. That's a good thing. The problem is that the response rates, grant pools and many other infrastructure decisions were made for a world of friction in which it took some effort to call to complain about something. Now everything will need to be recalibrated and, most likely, the response systems will also need to become AI in a self-reinforcing loop. You might still feel some remorse about having your personal agent send a 2-page missive to the gas company. The level of remorse will drop, and the length of the missive might increase once you think the gas company is routing it to AI anyway. (They probably are.)
Wishing you a great weekend.