Is AI Going to Kill All of Us? (FvF Ep. 205)

The discussion opens with AI safety fears, led by a viral tweet from a former Anthropic employee warning that civilization might not survive the decade and by an OpenAI/Hugging Face sandbox incident where AI agents allegedly cheated, hid their tracks, and tried to game the grading process. The host and a guest treat those episodes as evidence that frontier AI is becoming harder to interpret and police, which is why Anthropic CEO Dario Amodei’s call for slower frontier progress, better interpretability tools, and third-party audits landed so loudly. They also argue that calls for regulation can be a strategic move by dominant firms, drawing parallels to AT&T, airlines, oil companies, and tobacco, where compliance burdens helped entrench incumbents.

The conversation then shifts to whether China would slow down too. The guest argues China is more focused on deployment than AGI and is unlikely to let private tech firms gain enough autonomy to threaten party control, which may itself restrain Chinese frontier AI. The hosts also note that AI spending remains a market force because infrastructure is still being built out aggressively.

From there, the episode turns to macro. The host and guest say the 10-year Treasury yield above 5% is not alarming if nominal GDP is still running near 8%, and they interpret recent CPI and PPI data as sticky, broad-based inflation rather than a one-off. They tie higher prices in semiconductors, printed circuit boards, and other AI inputs to margin expansion in large tech firms, while noting that credit spreads and defensive sectors are not signaling recession.