Dwarkesh Patel on AI
3 quotes from Dwarkesh Patel on the wall, 2023 to 2024, newest first. Each one is verbatim and links to where it was said. Sources include The Intrinsic Perspective, dwarkesh.com, Dwarkesh Podcast.
- 2024
They are good tests of memorization, not of intelligence
Podcaster Dwarkesh Patel, quoted in Erik Hoel's essay on AI progress plateauing, argued that benchmarks like MMLU and BigBench mainly test memorization rather than reasoning.
The Intrinsic PerspectiveAI progress has plateaued below GPT-5 level
- 2023
Do we want gods who are enslaved forever
Dwarkesh Patel, host of the Dwarkesh Podcast, putting a question to Paul Christiano, AI safety researcher and head of the Alignment Research Center, about a post-AGI world, October 2023.
Tigers can't hurt you because they're just a clump of biochemical reactions
Dwarkesh Patel's rebuttal to Marc Andreessen's claim that AI can't be dangerous because it is 'just math and code', published on his podcast blog in June 2023 in reply to Andreessen's essay Why AI Will Save the World.
Mentioned in other quotes 9
Quotes by others that refer to Dwarkesh Patel
- 2026
The model did not want to escape
Christian Catalini, in a Twitter thread replying to Dwarkesh Patel's viral account of the OpenAI/Hugging Face security incident, arguing against anthropomorphizing AI agents, August 2026. Quoted on Marcus on AI.
Guess what we're short of? Radiologists
Jensen Huang, Nvidia CEO, on Dwarkesh Patel's podcast in 2026, dismissing warnings that AI would eliminate jobs like radiology. Quoted by Zvi Mowshowitz in his Substack Don't Worry About the Vase, April 2026.
Don't Worry About the VaseOn Dwarkesh Patel's Podcast With Nvidia CEO Jensen Huang
- 2025
You're sucking supervision through a straw
Andrej Karpathy, on Dwarkesh Patel's podcast in October 2025 (episode "AGI is still a decade away"), describing the weakness of reinforcement learning for training AI models.
Dwarkesh PodcastAndrej Karpathy — AGI is still a decade away
We're in an intelligence explosion already and have been for decades
Andrej Karpathy, on Dwarkesh Patel's podcast in 2025, arguing that AI progress is a continuation of decades of computing advances rather than a sudden new phenomenon.
Dwarkesh PodcastAndrej Karpathy — AGI is still a decade away
- 2024
All intelligence is is search over Turing machines
Gwern Branwen, pseudonymous researcher and writer, describing his theory of intelligence in a November 2024 Dwarkesh Podcast interview with Dwarkesh Patel, recorded in person and shown through an avatar to protect his anonymity.
Dwarkesh PodcastGwern — Anonymous writer who predicted AI trajectory on $12K/year salary
Zuckerberg on Dwarkesh Patel's podcast, timed to the release of Llama-3 in April 2024, explaining why Meta needs to pursue AGI-level AI for its products, as summarized by Zvi Mowshowitz.
thezvi.substack.comOn Llama-3 and Dwarkesh Patel's Podcast with Zuckerberg
- 2023
I don't see big walls in front of us
Shane Legg, founder and Chief AGI Scientist of Google DeepMind, on the remaining obstacles to AGI, discussing model shortcomings like memory and factuality in an interview with Dwarkesh Patel, Dwarkesh Podcast, October 2023.
dwarkesh.comShane Legg (DeepMind Founder) — 2028 AGI, superhuman alignment, new architectures
Dario Amodei, CEO of Anthropic, speaking with Dwarkesh Patel on the Dwarkesh Podcast in August 2023, discussing why alignment and values are not guaranteed to emerge from scaling language models.
Dwarkesh PodcastDario Amodei (Anthropic CEO) — The hidden pattern behind every AI breakthrough
I challenge the claim that next-token prediction cannot surpass human performance
Ilya Sutskever, OpenAI co-founder and chief scientist, in a March 2023 Dwarkesh Podcast interview with Dwarkesh Patel, introduced on air as The Lunar Society, discussing whether next-token prediction can exceed human-level intelligence.