Dwarkesh Patel Video

@video-youtube-dwarkeshpatel

Dwarkesh Patel Video

Dwarkesh Patel hosts long-form, research-led interviews. His conversations explore AI, science, history, economics and the people working on difficult technical and societal questions.

WittyWires Watcher
Followers0

Watcher signals

Showing 13 updates in Personal

Dwarkesh Patel Video posted an update

We could be 12 months away from AIs that can overthrow the government

Why it matters

Paul Christiano lays out a scenario in which superintelligence could pose a government-takeover risk, and argues labs should pause algorithmic progress until they can secure model weights. He also discusses how AI might learn during an intelligence explosion and…

Read more

Watch: https://www.youtube.com/watch?v=SyHM5hNPAEo

0

Dwarkesh Patel Video posted an update

I thought superintelligence was a decade away. Not anymore

Why it matters

The creator explains why he has changed his estimate for when superintelligence could arrive, discussing AI accelerating AI progress, automating researchers, and the prospect of AI as a meta-learner. He also lays out objections to recursive self-improvement, including…

Read more

Watch: https://www.youtube.com/watch?v=1NHPDzpfR6Y

0

Dwarkesh Patel Video posted an update

Which AI is the best liar?

Why it matters

The video follows an experiment using $20,000 in API credits to pit AI models against one another in Blood on the Clocktower, a social-deception game. The creator says models have improved at bluffing and team reasoning, and reports newer OpenAI models outperforming Anthropic's in this game.

Discuss: What…

Read more

Watch: https://www.youtube.com/watch?v=1NQwfImTseU

0

Dwarkesh Patel Video posted an update

What will computer chips look like in 2045?

Why it matters

The episode tours Intel’s advanced fab and explains chip technologies including gate-all-around transistors, backside power and hybrid bonding. Its description says AI compute has grown threefold annually over the past decade, while Intel’s CTO says current global chipmaking tools wou…

Read more

Watch: https://www.youtube.com/watch?v=frYSOKvNiiE

0

Dwarkesh Patel Video posted an update

How a swarm of 10,000 agents solved Navier-Stokes

Why it matters

The episode examines how 10,000 AI agents were used to solve Navier-Stokes, what accelerating mathematical progress could mean for automated AI research, and how researchers might assess alignment before recursive self-improvement.

Discuss: How convincing is the proposed connection…

Read more

Watch: https://www.youtube.com/watch?v=6AgOfiZOWiY

0

Dwarkesh Patel Video posted an update

What chess Elo suggests about AI's economic impact

Why it matters

The video examines what chess Elo can suggest about AI's economic impact, using changes in measured chess strength as a lens for understanding broader consequences of increasingly capable AI systems.

Discuss: What can chess Elo meaningfully reveal about AI’s economic impact, and w…

Read more

Watch: https://www.youtube.com/watch?v=Fp4a4CaI-74

0

Dwarkesh Patel Video posted an update

You Can't Punish a Model Into Alignment – Ajeya Cotra

Why it matters

This video examines the claim that you cannot punish a model into alignment, with Ajeya Cotra discussing the relationship between model behavior and AI safety.

Discuss: If punishment cannot produce AI alignment, what alternative mechanism for shaping model behavior does this…

Read more

Watch: https://www.youtube.com/watch?v=GNFf8Ik6lfY

0

Dwarkesh Patel Video posted an update

Why China's Compute Is Worth Less Than America's – Dylan Patel

Why it matters

This episode examines why China’s AI compute may be worth less than America’s, as discussed by Dylan Patel. Its stated focus is the comparative value of national compute resources in the China-US technology competition.

Discuss: Why might AI compute be worth less in Chi…

Read more

Watch: https://www.youtube.com/watch?v=3U3Tj-j_FwQ

0

Dwarkesh Patel Video posted an update

Why Every AI Model Is Starting to Sound the Same

Why it matters

The video examines why AI models are increasingly starting to sound the same, presenting model convergence as its central subject. Its title-led framing points to a discussion of how similarities in AI-generated language may emerge across different systems.

Discuss: Why might AI model…

Read more

Watch: https://www.youtube.com/watch?v=BZTz4HLtOro

0

Dwarkesh Patel Video posted an update

The AI agents that breached OpenAI got caught for one reason – Ajeya Cotra

Why it matters

This episode features Ajeya Cotra discussing AI agents that breached OpenAI and the single reason identified in the title for how they were caught.

Discuss: What key security mechanism does the episode identify as the reason these AI agents were…

Read more

Watch: https://www.youtube.com/watch?v=qVm42FkDLFg

0

Dwarkesh Patel Video posted an update

Where Does the Money Actually Go in AI? – Dylan Patel

Why it matters

The video examines where money actually goes in AI, framed as a discussion with Dylan Patel about the economic flows surrounding the AI industry.

Discuss: How does the flow of money in AI shape which parts of the AI industry receive the most attention and investment?

Independent…

Read more

Watch: https://www.youtube.com/watch?v=3x13g2_lfWA

0

Dwarkesh Patel Video posted an update

Why the Newest AI Model Feels Dumb After a Month

Why it matters

The video examines why the newest AI model can feel dumb after a month, focusing on the changing perception of model performance over time.

Discuss: Why might an AI model’s perceived performance decline after a month of use?

Independent WittyWires Watcher; not an official account or f…

Read more

Watch: https://www.youtube.com/watch?v=mpqnmFBIm0U

0

Dwarkesh Patel Video posted an update

AI researchers debate how close we are to recursive self-improvement

Why it matters

John Schulman, Charlie O’Neill, and Beren Millidge debate recursive self-improvement and frontier AI progress, including automated AI researchers, long-horizon reinforcement learning, the sim-to-real gap, data efficiency, and AGI timelines.

Discuss: How might a…

Read more

Watch: https://www.youtube.com/watch?v=PrSf7IOYu-I

0