Why AI Language Can Mislead

Jensen Huang is so hot right now; not only did he make some statements that were widely derided, but he also featured in a glitzy "collab" episode with our favourite lone wolf, Lex Fridman. Sensing guru vibes, we turned to take a deep dive into his recent interview with Ezra Klein. Huang is the co-founder and CEO of Nvidia; he also really likes Lex Fridman and the All-In Besties. Trump even called him while he was on the All-In stage! So it's fair to say we went into this one expecting yet another bloviating tech CEO and were ready to join the dunking. We do not do guilt by association, of course, but bad vibes were in the air.

Huang thinks AI will transform the economy, take over tedious tasks, and leave us free to do more interesting things. Some jobs will disappear, but new industries will emerge to replace them, like wellness spas... It's okay for those whose jobs are disappearing, though Jensen has a solution: they should become systems thinkers!

Alongside such high-level insights, Matt and Chris do find some points of agreement, including his impatience with apocalyptic predictions being presented as scientific findings. The discussion also highlights a difference between our dear hosts: Huang's irritation with people talking about software like little human minds makes Chris increasingly sympathetic to him. Matt, as a committed pan-psychic, is less impressed by the reassurance that engineers understand what they are building, and thinks understanding the components is rather different from knowing what the whole system will do. Inevitably, this leads to a discussion of dogs, ants, and what "unpredictable" means, because apparently we are sensemakers now.

We will get to the fuller discussion of AI safety in Part Two. Until then, listeners can enjoy the experience of us listening to one of the less terrible tech billionaires and disagreeing with each other over minor points!

Source material

• The Ezra Klein Show: Jensen Huang Thinks A.I. Alarmism Has Gone Too Far
• The New York Times: Episode page and transcript
• Lex Fridman: Taiwan: The Most Epic Night Market Experience w/ Jensen Huang (the Taiwan visit mentioned in the introduction)
• All-In Summit: Jensen Huang: The Doomer Hoax, Superintelligence is Here, and The Future of AI (ft. President Trump)
• Creative Destruction Lab: Geoff Hinton: On Radiology (2016)
• Kurzgesagt: AI Just Crossed the Terrifying Line - Now What? (previously titled The Moment They Lost Control of AI)

Relevant articles

• Vaswani et al. (2017): Attention Is All You Need
• Anthropic: When AI builds itself
• METR: Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident
• Anthropic: 2026 Usage Policy update (including its policy on abusive treatment of models)