Y Combinator 邀 Jacob Coxon 和 Robin Hanson 辯論政府該不該放慢 AI 發展
Should governments slow frontier AI? Jacob Coxon and Robin Hanson debate [video]
AI 發展可能帶來哪些風險,政府該不該出手放慢腳步?
- Y Combinator 邀請 Jacob Coxon 和 Robin Hanson 辯論。Coxon 提出超級智慧可能很快出現,也談到模型行為難以理解和風險難以補償。
- Hanson 則質疑政府管制能否奏效,主張考慮以責任追究等方式處理。兩人也討論 AI 對工作、監管機構和安全的影響,並在最後回應觀眾提問。
AI 伴隨問答
針對這篇報導有疑問?可直接點選快捷問題或向 AI 追問背景脈絡與技術細節:
Jacob Coxon vs. Robin Hanson: Should We Slow Down AI? - YouTube
Back [](https://www.youtube.com/ "YouTube Home") KR
Skip navigation
Search
Search with your voice
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
[](https://www.youtube.com/ "YouTube Home") KR
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
Jacob Coxon vs. Robin Hanson: Should We Slow Down AI?
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
Tap to unmute
2x
Jacob Coxon vs. Robin Hanson: Should We Slow Down AI?
Y Combinator 24,387 views 17h ago
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
Copy link
Info
Shopping
If playback doesn't begin shortly, try restarting your device.
•
You're signed out
Videos you watch may be added to the TV's watch history and influence TV recommendations. To avoid this, cancel and sign in to YouTube on your computer.
Cancel Confirm
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
Share
[](https://www.youtube.com/watch?v=lwfewYL7X0o "Share link")- [x] Include playlist
An error occurred while retrieving sharing information. Please try again later.
0:00
[](https://www.youtube.com/watch?v=lwfewYL7X0o)[](https://www.youtube.com/watch?v=96jN2OCOfLs)
0:00 / 0:00
Live
•Watch full video
•
YC Debate Topic: Should We Slow Down AI?
•
46:58 Gemini and the path to AGI with Koray Kavukcuoglu Google DeepMind 42K • 12h ago Live Playlist ()Mix (50+)38:49 Nick Bostrom on Superintelligence, Longer Lives, and Deep Utopia a16z 12K • 17h ago Live Playlist ()Mix (50+)45:28 Can Canada's Francophone Network Unlock New Markets? | Toronto Global Forum 2026 IEFA TV 9 • 8h ago Live Playlist ()Mix (50+)27:07 Geoffrey Hinton | Full Episode 11.13.26 | Firing Line with Margaret Hoover | PBS Firing Line with Margaret Hoover | PBS 265K • 6d ago Live Playlist ()Mix (50+)39:30 What OpenAI Did After the Hugging Face Hack The New York Times 109K • Streamed 1d ago Live Playlist ()Mix (50+)35:39 Can Zohran Mamdani Fix the Cost of Living in New York City?The New York Times 317K • Streamed 1d ago Live Playlist ()Mix (50+)13:46 java situation is insane Low Level 528K • 16h ago Live Playlist ()Mix (50+)1:00:06 The Rise Of AI Coworkers Y Combinator 167K • 5d ago Live Playlist ()Mix (50+)14:41 The making of Anthropic Anthropic 121K • 11h ago Live Playlist ()Mix (50+)55:04 Elon Musk on Why the Future Should Make You Glad to Be Alive | TomorrowX Summit iConnections 253K • 1d ago Live Playlist ()Mix (50+)35:44 Anthropic C.E.O.: Massive A.I. Spending Could Haunt Some Companies The New York Times and New York Times Events 448K • Streamed 11mo ago Live Playlist ()Mix (50+)29:49 Andrej Karpathy: From Vibe Coding to Agentic Engineering w/ Stephanie Zhan Sequoia Capital 1.7M • 6mo ago Live Playlist ()Mix (50+)
Sign in to confirm you’re not a bot
This helps protect our community
Jacob Coxon vs. Robin Hanson: Should We Slow Down AI?
Y Combinator
2.44M subscribers
Subscribe
Subscribed
239
Download
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
K
views
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
h ago
24,387 views • Nov 19, 2026
Should governments intervene to substantially slow the progress of frontier AI? In the inaugural YC debate, Jacob Coxon, an ex-Anthropic and OpenAI researcher, faces off against Robin Hanson, an economist at GMU, about AI takeoff and what government regulation should (or shouldn’t) do about it.…more
more
How this was made
Auto-dubbed
Audio tracks for some languages were automatically generated. Learn more
Chapters
View all
### YC Debate Topic: Should We Slow Down AI? ### YC Debate Topic: Should We Slow Down AI? 0:00
YC Debate Topic: Should We Slow Down AI?
0:00
### Jacob's case: Superintelligence by 2028 ### Jacob's case: Superintelligence by 2028 2:42
Jacob's case: Superintelligence by 2028
2:42
Models already behaving in ways we don't understand
7:02
### Robin's case: the danger of abstraction ### Robin's case: the danger of abstraction 9:18
Robin's case: the danger of abstraction
9:18
Why new technologies take decades to diffuse
11:22
### Liability instead of regulation ### Liability instead of regulation 13:12
Liability instead of regulation
13:12
Why regulating AI is different from regulating drugs
14:50
Jacob's rebuttal: general intelligence is one number
16:00
Transcript
Follow along using the transcript.
Show transcript
### Y Combinator 2.44M subscribers
Video details
Date
Nov 19, 2026
Views
24,387
Likes
239
Show less
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
# Jacob Coxon vs. Robin Hanson: Should We Slow Down AI?
24,387 views
Nov 19, 2026
239
Download
97 Comments
Sort comments
Sort by
Top Show featured commentsNewest Show recent comments, including potential spam
Add a comment...
@SuthaKamal-h9o
Hanson’s assertion that LLMs are descendants of humans is just absurd nonsense, and it’s a shame he wasn’t thoroughly taken apart for such a silly suggestion. Habibi, the LLM doesn’t exist between clock ticks of the von Neumann machines they run on.
Show less Read more
Like
7
Dislike
Reply
@danielsierrala
Why is Robin Hanson in this debate? It would have been nice to have two experts on the subject debating
Show less Read more
Like
22
Dislike
Reply
3 replies
Hide replies
3 replies
@ZanesZoneUnlimited
Am i the only one that wanted to scream at Hanson and say "but they arent human!!!" Bro they are not human children. Why does that even need to be said...
Show less Read more
Like
11
Dislike
Reply
5 replies
Hide replies
5 replies
@Alverin
Abstractions don't JUST drop detail, ideally, they prioritize the details that matter and then they discard the details that don't. As an example, a map is an abstraction of the landscape that drops the details of the color of things in that landscape and replaces it with more detailed information about where structures are located. Including too many details hurts your ability to reason about things depending on what you're trying to reason about (such as, how do I get downtown?)
Show less Read more
Like
2
Dislike
Reply
@jayinjapan
I will say I don’t think we actually understand how much AI is actually effectively creating improvements across the economy. Every thought leader, investment banker, and influencer is compromised and their opinions aren’t reliable. It’s true to say that there’s a lot of jobs that could be replaced with an LLM, it’s also true to say that many companies fired people claiming it was because of AI but it was just cutting costs, it’s also true to say that some jobs that were eliminated were not needed in the first place, a lot of CEOs are also forced to say that they made lots of changes just to appear like they are AI forward or whatever.
Show less Read more
Like
3
Dislike
Reply
@NewtonBastiat
1:02:30 Robin snuck in the phrase "per incident", but surely the covariance is so high that $100B per incident would need orders of magnitude greater coverage?
Show less Read more
Like
1
Dislike
Reply
@Bubbabrine
I'm baffled by Robin Hanson's complete lack of understanding of this topic. "Regulation is hard, so we shouldn't even try" is the most naive/ignorant stance possible. The magnitude of the pace of AI development compared to anything that's ever happened in human history has no equal. AI is not our children, AI is alien. He has no comprehension of this topic.
Show less Read more
Like
10
Dislike
Reply
@kamaaina7183
By the time humans wait for a massive incident that insurance companies realize uninsurable, it’s too late. Too late in an absolute thresh hold,and no turning back. Once AI gains physical presence, they will supersede humans on the earth with logics and super intelligence we are no much. Combine it with competition over energy requirements for both species, the results are obvious.
Show less Read more
Like
2
Dislike
Reply
1 reply
Hide replies
1 reply
@aronbuch4275
no offense, but i feel like robin hanson is completely unfamiliar with the alignment problem. jacob lays out an argument for why AI will likely soon be superhuman and we won't be able to control them robin replies "eh, that sounds kinda crazy, so we're probably fine" ...okay???
Show less Read more
Like
28
Dislike
Reply
6 replies
Hide replies
6 replies
@Alverin
When Robert Hanson says the AI in the Hugging Face incident were pursuing the rewards we gave them, the issue is that you can't just have a reward, you have to have a reward and safe path to pursue it, if you just have a reward and let the AI do whatever it wants to achieve it, at a certain level of sophistication it will just edit you (the auditor) out of the equation, which is exactly what the Hugging Face incident showed. The fact that he immediately followed his claim up with "we are making AI's in our image" really gives me populist vibes from him lol
Show less Read more
Like
1
Dislike
Reply
@jayinjapan
Hanson hasn’t spent 20 minutes on LinkedIn or any serious organisation it seems in the last 3 years. Leaders are pushing for every employee to use nothing but AI without verifying what’s going on, and people don’t bother understanding how it works and what it’s doing for every task. That’s the whole point. Don’t think. Don’t check. Don’t look. Just release.
Show less Read more
Like
1
Dislike
Reply
@SuthaKamal-h9o
45:02 re weapons: Sorry Hanson, physical weapons can be locked in buildings. The NSA can’t secure a lot. You are committing one category error after another.
Show less Read more
Like
2
Dislike
Reply
@jayinjapan
Hanson is just a bad-faith debater. His counter-argument to the pushback on the children analogy was “so you don’t want your kids to be smarter than you because it bruises your ego?”. Absolutely unhinged. AI aren’t human, they have none of the cooperative and collaborative incentives that humans do, or the species kinship. If they think that it makes sense to burn all humans for fuel to power data centers so it can live, it wouldn’t have qualms about it. Even if it didn’t inherit our culture, we’ve seen what happens when a person gets ultimate power and doesn’t depend on its people to get rich. They are called dictators. Except these ones wouldn’t need to eat/sleep/get sick/die, or have minions to do their bidding, etc.
Show less Read more
Like
6
Dislike
Reply
@TheST8Mverse
How does someone with no digital footprint gain 175 million views on his first interaction on X
Show less Read more
Like
1
Dislike
Reply
@PabloMejía-k7f
I came in hoping to find some appeasement in hanson’s arguments but he has no idea what he’s talking about.
Show less Read more
Like
1
Dislike
Reply
@s.sasquatch1789
We wanna sprend collaboration as the main driver of institutional development and security. speciation is best scenario in my opinion. everyone gets an agentic! personaly developed executive assistant. individual motivation can be withstood via collaborative institutions to preserve and utilze cultural understanding.
Show less Read more
Like
Dislike
Reply
@CREEPYASMRREAPER
Love Jacob, watch all his stuff, and Conor Leahy, who's been warning for years!
Show less Read more
Like
2
Dislike
Reply
1 reply
Hide replies
1 reply
@Zelariath
Part of me thinks hes only argueing this becuase he needs AI to accelerate to accelerate healthcare so he doesn't die of old age in the next couple decades. It seems like he would risk killing everyone to save his own life.
Show less Read more
Like
15
Dislike
Reply
1 reply
Hide replies
1 reply
@SuthaKamal-h9o
Re Hanson’s comment about regulatory capture: sure, it’s not an argument against regulation. Homeboy seems so very clueless about what he’s speaking about. He’s got a few pithy sound bites that seem thoughtful at first glance.
Show less Read more
Like
1
Dislike
Reply
@s.sasquatch1789
9:27 my two factor is abstraction and collaboration as replicators.
Show less Read more
Like
Dislike
Reply
3 replies
Hide replies
3 replies
Comments 97
Top Show featured commentsNewest Show recent comments, including potential spam
In this video
Chapters
Transcript
Chapters
### YC Debate Topic: Should We Slow Down AI? ### YC Debate Topic: Should We Slow Down AI? 0:00
YC Debate Topic: Should We Slow Down AI?
0:00
### Jacob's case: Superintelligence by 2028 ### Jacob's case: Superintelligence by 2028 2:42
Jacob's case: Superintelligence by 2028
2:42
Models already behaving in ways we don't understand
7:02
### Robin's case: the danger of abstraction ### Robin's case: the danger of abstraction 9:18
Robin's case: the danger of abstraction
9:18
Why new technologies take decades to diffuse
11:22
### Liability instead of regulation ### Liability instead of regulation 13:12
Liability instead of regulation
13:12
Why regulating AI is different from regulating drugs
14:50
Jacob's rebuttal: general intelligence is one number
16:00
Why normal liability won't cover extinction risk
17:36
Robin's rebuttal: regulating this means regulating all compute
19:52
### AI as our descendants ### AI as our descendants 21:44
AI as our descendants
21:44
### Diminishing returns ### Diminishing returns 23:38
Diminishing returns
23:38
Direct exchange: is it just the rate of change?
24:40
How long until most cognitive work is automated?
26:48
### The jobs we have no data for ### The jobs we have no data for 29:21
The jobs we have no data for
29:21
### How do you align the regulators? ### How do you align the regulators? 32:45
How do you align the regulators?
32:45
Reward hacking, and what Hugging Face actually showed
35:03
### Why not an equilibrium of AIs watching AIs ### Why not an equilibrium of AIs watching AIs 36:11
Why not an equilibrium of AIs watching AIs
36:11
Q&A: what regulation would you actually want?
39:50
### Q&A: what about China? ### Q&A: what about China? 41:36
Q&A: what about China?
41:36
### Q&A: an FAA for AI labs? ### Q&A: an FAA for AI labs? 42:58
Q&A: an FAA for AI labs?
42:58
Q&A: the military is already deploying this
43:45
Q&A: does regulation just move the reward hacking up a layer?
47:46
Q&A: what unemployment rate can an economy withstand?
51:27
### Q&A: what's the downside of a pause? ### Q&A: what's the downside of a pause? 56:25
Q&A: what's the downside of a pause?
56:25
### Q&A: what would change your mind? ### Q&A: what would change your mind? 58:09
Q&A: what would change your mind?
58:09
### Closing statements ### Closing statements 1:02:01
Closing statements
1:02:01
### The final audience vote ### The final audience vote 1:04:24
The final audience vote
1:04:24
Sync to video time
Jacob Coxon vs. Robin Hanson: Should We Slow Down AI?
Y Combinator
Y Combinator
239 Likes
24K Views
17h Ago
Should governments intervene to substantially slow the progress of frontier AI? In the inaugural YC debate, Jacob Coxon, an ex-Anthropic and OpenAI researcher, faces off against Robin Hanson, an economist at GMU, about AI takeoff and what government regulation should (or shouldn’t) do about it. Chapters: 00:00 - YC Debate Topic: Should We Slow Down AI? 02:42 - Jacob's case: Superintelligence by 2028 07:02 - Models already behaving in ways we don't understand 09:18 - Robin's case: the danger of abstraction 11:22 - Why new technologies take decades to diffuse 13:12 - Liability instead of regulation 14:50 - Why regulating AI is different from regulating drugs 16:00 - Jacob's rebuttal: general intelligence is one number 17:36 - Why normal liability won't cover extinction risk 19:52 - Robin's rebuttal: regulating this means regulating all compute 21:44 - AI as our descendants 23:38 - Diminishing returns 24:40 - Direct exchange: is it just the rate of change? 26:48 - How long until most cognitive work is automated? 29:21 - The jobs we have no data for 32:45 - How do you align the regulators? 35:03 - Reward hacking, and what Hugging Face actually showed 36:11 - Why not an equilibrium of AIs watching AIs 39:50 - Q&A: what regulation would you actually want? 41:36 - Q&A: what about China? 42:58 - Q&A: an FAA for AI labs? 43:45 - Q&A: the military is already deploying this 47:46 - Q&A: does regulation just move the reward hacking up a layer? 51:27 - Q&A: what unemployment rate can an economy withstand? 56:25 - Q&A: what's the downside of a pause? 58:09 - Q&A: what would change your mind? 1:02:01 - Closing statements 1:04:24 - The final audience vote Apply to Y Combinator: https://www.ycombinator.com/apply Work at a startup: https://www.ycombinator.com/jobs…...more
...more Show less
How this was made
Auto-dubbed
Audio tracks for some languages were automatically generated. Learn more
Chapters
View all
### YC Debate Topic: Should We Slow Down AI? ### YC Debate Topic: Should We Slow Down AI? 0:00
YC Debate Topic: Should We Slow Down AI?
0:00
### Jacob's case: Superintelligence by 2028 ### Jacob's case: Superintelligence by 2028 2:42
Jacob's case: Superintelligence by 2028
2:42
Models already behaving in ways we don't understand
7:02
### Robin's case: the danger of abstraction ### Robin's case: the danger of abstraction 9:18
Robin's case: the danger of abstraction
9:18
Transcript
Follow along using the transcript.
Show transcript
### Y Combinator 2.44M subscribers
Video details
Date
Nov 19, 2026
Views
24,387
Likes
239
Transcript
NaN / NaN
Gemini and the path to AGI with Koray Kavukcuoglu
Google DeepMind
42K 12h ago
Nick Bostrom on Superintelligence, Longer Lives, and Deep Utopia
a16z
12K 17h ago
Can Canada's Francophone Network Unlock New Markets? | Toronto Global Forum 2026
IEFA TV
9 8h ago
Geoffrey Hinton | Full Episode 11.13.26 | Firing Line with Margaret Hoover | PBS
Firing Line with Margaret Hoover | PBS
265K 6d ago
What OpenAI Did After the Hugging Face Hack
The New York Times
109K Streamed 1d ago
Can Zohran Mamdani Fix the Cost of Living in New York City?
The New York Times
317K Streamed 1d ago
java situation is insane
Low Level
528K 16h ago
The Rise Of AI Coworkers
Y Combinator
167K 5d ago
The making of Anthropic
Anthropic
121K 11h ago
Elon Musk on Why the Future Should Make You Glad to Be Alive | TomorrowX Summit
iConnections
253K 1d ago
Anthropic C.E.O.: Massive A.I. Spending Could Haunt Some Companies
The New York Times and New York Times Events
448K Streamed 11mo ago
Andrej Karpathy: From Vibe Coding to Agentic Engineering w/ Stephanie Zhan
Sequoia Capital
1.7M 6mo ago
Palmer Luckey on War 2.0
New York Times Events
155K 14h ago
Ramp’s Bet on the 100x Employee
Y Combinator
24K 1d ago
Kai-Fu Lee on the Global AI Race | The Mishal Husain Show
Bloomberg Podcasts and Bloomberg Television
551K 2mo ago
Sam Altman: ‘It’s Hard To Imagine the AI Revolution Without Jensen’ | Silicon Valley & The World
Semafor
90K 3d ago
Training Sand to Think: Artificial General Intelligence & Future of Physics
Perimeter Institute for Theoretical Physics
355K 5mo ago
AI Memorized the Internet. That's Not Intelligence. | Yann LeCun
Futurology and Berggruen Institute
118K 2d ago
Anthropic IPOs, GPT 6.2 Astra Arrives, But There’s a Bigger Story
AI Explained
83K 12h ago
If AI Can Do the Math, What Do the Mathematicians Do? (ft. a16z Researchers)
a16z Frontier
11K 1d ago
Show more
Show more
[](https://www.youtube.com/watch?v=lwfewYL7X0o)
46:58 New
38:49 New
45:28 New
27:07 New
39:30 New
35:39 New
13:46 New
1:00:06 New
14:41 New
55:04 New
35:44