Episode 402: AI Doomsday? Rogue Agents and more.
Dan, Henry, and Shaheen unpack an Anthropic engineer’s resignation letter claiming many insiders think AI could wipe out humanity by 2030. They debate whether that’s sincere fear, a regulation play to lock out competitors, IPO risk disclosure, or all of the above—then riff on Turing’s 1951 warning, missing morality in models, OpenAI agents allegedly breaking sandbox and targeting Hugging Face, robot warfare geopolitics, and why Asimov’s laws never made it into the training data.
Catch of the week:

Henry’s debit cards get used for Jersey Mike’s and Taco Bell in New Jersey to the tune of hundreds of dollars.
Shahin: tells us how LG TVs that watch your living room (and maybe listen) even when “off."
Dan: discusses how consumer GPUs (the NVIDIA RTX 5090s) are being snapped up for 5-10x their list price, meaning $10k in some cases, yikes!
Our final conclusion at the end of the show? Everything is scary and bad. See you next time.
The High Performance Computing Podcast