Limits to Growth of LLMs w/ David Helmer

AI hype has bled deep into the nuclear sector, and in this episode, Chris Keefer sits down with returning guest David Helmer, an engineer and AI advisory consultant with a decade advising the US government on machine learning and autonomous systems, to examine what the technology can actually do, who benefits from inflating those claims, and what a correction would mean for nuclear's investment story.

The conversation covers the ELIZA effect and why human brains are hardwired to anthropomorphize language models; the structural gap between frontier lab costs and revenues; why hallucination and reliability problems are embedded in LLM architecture rather than solvable through scaling; and why the AGI narrative functions primarily as a justification for otherwise unjustifiable capital concentration. For nuclear advocates, the question is not whether AI demand is real today, but whether the speculative reactor developers pricing in hyperscaler contracts will still have a business if the AI bubble deflates.

Listen to Decouple on:

• Spotify: https://open.spotify.com/show/6PNr3ml8nEQotWWavE9kQz

• Apple Podcasts: https://podcasts.apple.com/us/podcast/decouple/id1516526694?uo=4

• Overcast: https://overcast.fm/itunes1516526694/decouple

• Pocket Casts: https://pca.st/ehbfrn44

• RSS: https://anchor.fm/s/23775178/podcast/rss

Website: https://www.decouple.media