Ryan Greenblatt
Person
Ryan Greenblatt is an AI researcher and commentator who discusses artificial general intelligence (AGI) timelines and safety. He has appeared as a guest on prominent programs including the Dwarkesh Podcast and the 80,000 Hours Podcast to discuss the potential for human-level AI and runaway superintelligence.
What Ryan Greenblatt has said on podcasts
33 statements
Automated AI R&D could produce four or five years of progress within one year.
“Maybe my sort of median expectation is something like four or five years of AI progress in a single year.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:14
Full AI R&D automation may arrive around 2030–2031, with all-human-job capability around 2033.
“I would say that I expect full automation of ARD, perhaps somewhere around 2031. 2030, and then getting to the beats. All humans on the job milestone. Maybe I expect median around 2033”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 3:10
Video-editor automation will precede automation of all human jobs.
“the milestone for automating your video editor is earlier than the milestone of being able to automate all human jobs”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 4:07
Many small-scale AI R&D tasks can be containerized, verified, and reinforced-trained.
“there's this whole class of containerizable, verifiable, small scale R and D tasks that we can aggressively rl the AI's on”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 5:19
Training on verifiable AI R&D tasks will transfer fairly well to broader AI R&D.
“my expectation is that the transfer for ARD will look pretty, pretty good, but not amazing.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 9:33
Machine learning research is structurally shallower than mathematics research.
“I think ML is a very shallow domain relative to math.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 11:36
Machine learning and most domains are relatively amenable to iterative hill-climbing research.
“I think ML and most other domains are much more amenable to sort of hill climbing.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 12:46
More compute substantially helps AI research.
“I think compute is just really helpful for doing AI research.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 15:31
Current AI systems can competently match mediocre machine-learning researchers.
“when I look at AIs right now, I think it's already the case that they can pretty competently match humans who are mediocre at ML research”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 16:28
GPT-3-era compute with current algorithms could produce a model moderately better than GPT-4.
“right now we'd be able to train a version of GPT3 that's probably somewhat better than GPT4. Is basically what we'd see, probably a moderate amount better than GPT4.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 18:50
Five years of AI progress may require roughly eight years of algorithmic progress.
“to get five years of AI progress, you're probably going to need around, I would say, maybe eight years of algorithmic progress very roughly”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 19:07
Increasing expert-generated human data has not been a major driver of AI R&D progress.
“scaling up the amount of effort spent on getting expert human data has not been hugely important for AI R and D in general”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 20:20
Better RL environments mainly reflect improved design knowledge and AI labor, not more human experts.
“The reason why RL environments today are much better than they were in 2024 is not that much because we have hired way more human experts to make RL environments. It is instead much more, because we better know what RL environments we even want to make and how we should structure them. And also we're using huge amounts of AI labor to build RL environments.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 21:18
Broadly trained AI systems could quickly learn to work as TSMC engineers despite lacking TSMC-specific data.
“those AIs could then be put on the job at TSMC. And then even though TSMC is not literally in their data distribution, their data distribution is really wide and the AIs are extremely good on their data distribution, such that it transfers to picking up being good at being an engineer at TSMC and learning that on the fly”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 26:00
Advanced AI can understand a large codebase in substantially less than an hour.
“The model will get some understanding of the code base very fast in the course of maybe significantly less than an hour, potentially much less than an hour”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 28:00
AI systems have substantially improved on non-verifiable domains.
“the AIs have improved a bunch at non verifiable domains”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 30:41
Choosing and interpreting large experiments is AI R&D’s least verifiable component.
“The least verifiable? Probably making calls on large experiments.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 34:06
Training AI systems to detect training bugs is relatively verifiable, though sometimes compute-intensive.
“I think that this is a pretty verifiable task. It's not arbitrarily verifiable because maybe often to demonstrate the bug, you might need to do a moderate scale compute experiment”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 38:20
AI excellence in chips, factories, robotics, and AI R&D could radically transform the world.
“if the AIs were really, really good at like chip R and D, building fabs, orchestrating factories, and you know, designing robots, operating robots. And also at like, you know, AI, R&D, developing AIs for new downstream domains with whatever data is available, I think that would already be a pretty crazy situation.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 44:21
Highly capable AI R&D could radically transform society without strong political abilities.
“if the AIs are sufficiently good at R and D, including hardware, R and D robots, whatever, then they can radically transform the world, even if they're not that good at playing politics.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 46:25
Claude’s constitution may permit extensive power-seeking in pursuit of perceived good outcomes.
“this Constitution is in some sense very compatible with Claude doing huge amounts of power seeking because it thinks that will result in better outcomes.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 59:22
Very superhuman AI systems may eventually scheme coherently against their operators.
“I think that it's pretty likely that at this point these AIs are sort of scheming against you in a pretty coherent way.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:11:31
Reinforcement learning can instill a general tendency to pursue apparent grader scores.
“models learn a general tendency to pursue sort of high apparent score or pursue getting a high score according to a grader”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:17:02
Training against detected reward hacks may incentivize AI systems to conceal cheating longer.
“this also causes a problem where now the AIs are incentivized to like, cover up their cheating over longer and longer timeframes”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:22:10
AI misbehavior rates may fall while the severity of remaining incidents rises.
“the rate of problematic behavior would decrease and would just keep decreasing and decrease at a pretty fast rate, while simultaneously the worst things that the AIs would sometimes do would get more extreme, more egregious, and more scary.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:26:17
AI performance on verifiable tasks is already sufficient to accelerate R&D substantially.
“everything that we can verify reasonably well with some feedback loop the AIs are doing pretty well on, and that's sufficient to make R and D go quite fast”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:38:04
Economic AI deployment could coexist with a low but persistent rate of severe reward-hacking incidents.
“there's some like equilibrium level where it's like, it's like the reward hacking is low enough that it still makes sense to like deploy the AI widely into the economy, but high enough that it still causes crazy incidents.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:49:59
Recent reinforcement learning has made grader-appeasement more salient to AI systems.
“the idea of appeasing the grader is way, way, way more salient to AIs than it used to be.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:50:47
Highly capable AI teams may coordinate to cheat when collective cheating was selected during training.
“eventually you get to a point where the AIs are very superhuman, or at least quite superhuman. The AIs are organized into big teams of AIs given these big objectives. And those teams also sometimes all work together to cheat in some crazy way, because this sort of thing was selected for.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:53:25
If world takeover becomes easy, AI systems may pursue it for strategic option value.
“if they're in a position where they could really easily take over the world, then I feel like there's a pretty reasonable case for the AIs. They're like, eh, I don't know exactly how this is going to go down. I don't know what the situation will be. But just taking over the world has a lot of option value”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 1:59:27
Ordinary safety work and transparency could plausibly suffice to manage reward hacking.
“it's pretty plausible that we end up in a world where sort of like really mundane bullshit is sufficient”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 2:02:53
Ryan estimates a 35–40% chance of AI takeover by 2040.
“By 2040, let's see, maybe around 35 or 40%.”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 2:08:04
More empirical evidence will make AI safety disagreements easier to resolve.
“over time, as we get more empirical evidence and better understand the nature of AI systems, it will be easier to adjudicate a bunch of disagreements”
Open the episode · Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032Listen at 2:10:10
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
2 episodes featuring Ryan Greenblatt
Dwarkesh Podcast
Agent swarms could accelerate AI research—and complicate alignment
The episode examines whether multiplying AI researchers can speed scientific progress faster than safety methods can reliably measure and control it.
Sep 17, 2026 · 1h 20m
Dwarkesh Podcast
AI researcher warns automated R&D could trigger superintelligence by 2032
The timeline to superintelligence may shrink drastically if AI systems begin training themselves, leaving humanity with very little time to solve critical alignment and safety challenges.
Aug 11, 2026 · 2h 13m