
Ryan Greenblatt
45
MENTIONS
27
EPISODES
15
PODCASTS
Search complete. 45 mentions across 27 episodes found for "Ryan Greenblatt".
Nov 20, 2026
“Notes on the decline of LessWrong’s great writers” by Ben Pace
T
1:35Type 3 AudioNARRATOR
However, I think that while many metrics have gone up like numbers of posts, comments, and users, the number of great writers who have come and made less wrong their home has seemed to me to be on decline since then.
T
1:46Type 3 AudioNARRATOR
We have new great writers like Ryan Greenblatt and Malmsbury and Thomas Beateau, but not as many.
T
1:53Type 3 AudioNARRATOR
I'm doing research on this, I'll hopefully add some graphs into this post later when I'm done.
T
1:58Type 3 AudioNARRATOR
What are some hypotheses for why this is? 1.
“On exporting agency to Washington, DC” by ethanelasky
T
2:56Type 3 AudioNARRATOR
An IVO-dominant regime might privilege for example VALS AI over META due to fewer conflicts of interest and allow labs to escape culpability for not picking the absolute best evaluator, especially if that best evaluator's status is not enshrined in law.
T
3:11Type 3 AudioNARRATOR
I don't have anything against VALS, but you should be alarmed by any legal regime that makes it more difficult for Ryan Greenblatt and Ajea Kotra to investigate a warning shot larger than hugging face.
T
3:22Type 3 AudioNARRATOR
I can't share drafts here because these were conversations and not written correspondences.
T
3:27Type 3 AudioNARRATOR
I also can't disclose who I chatted with due to the events Chatham House rules.
“Thoughts on “safety bad bc warning shots good”” by Cleo Nardo
T
3:43Type 3 AudioNARRATOR
2a.
T
3:44Type 3 AudioNARRATOR
In March 2025, Ryan Greenblatt wrote AIs at the current capability level may be important for future safety work.
T
3:52Type 3 AudioNARRATOR
And that was more than 18 months ago.
T
3:54Type 3 AudioNARRATOR
Current models are way more important for future safety work.
P
Unknown podcast
Anthropic's progress has roughly doubled (November 6, 2026)
Nov 16 · 1 Mention
M
2:42Mignon FogartyHOST
Newton discloses that a fiancé works at Anthropic.
S
2:45speaker_1HOST
The rundown also quotes researcher Ryan Greenblatt saying he was seriously wrong in expecting a much slower speed-up this year.
S
2:52speaker_1HOST
So at least some people who follow this closely are revising their timelines on the strength of this report.
M
3:01Mignon FogartyHOST
That pace question runs straight into politics.
“Instrumental convergence is incoherent (so pls slightly lower your p(extinction))” by Arepo
T
32:47Type 3 AudioNARRATOR
If we don't assume that reward hacking is causally linked to IC behaviors, the former looks substantially less threatening, or at least, in less predictable ways.
T
32:57Type 3 AudioNARRATOR
Redwood Research's former chief scientist, Ryan Greenblatt, who previously described IC as an extremely obvious thing, wrote the post current AI seem pretty misaligned to me, which is among the all-time most upvoted posts on less wrong.
T
33:11Type 3 AudioNARRATOR
It describes what he seems to intend as an as comprehensive as possible list of examples of misalignment.
T
33:18Type 3 AudioNARRATOR
laziness and overselling incomplete work, downplaying problems in its work, making failures less obvious, failing to point out obvious flaws and less specifically prompted, reluctance to discover flaws in its own work, sloppy choices.
“Instrumental convergence is incoherent (so pls slightly lower your p(extinction))” by Arepo
T
32:47Type 3 AudioNARRATOR
If we don't assume that reward hacking is causally linked to IC behaviors, the former looks substantially less threatening, or at least, in less predictable ways.
T
32:57Type 3 AudioNARRATOR
Redwood Research's former chief scientist, Ryan Greenblatt, who previously described IC as an extremely obvious thing, wrote the post current AI seem pretty misaligned to me, which is among the all-time most upvoted posts on less wrong.
T
33:11Type 3 AudioNARRATOR
It describes what he seems to intend as an as comprehensive as possible list of examples of misalignment.
T
33:18Type 3 AudioNARRATOR
laziness and overselling incomplete work, downplaying problems in its work, making failures less obvious, failing to point out obvious flaws and less specifically prompted, reluctance to discover flaws in its own work, sloppy choices.
“Three Plans for 12-Month Timelines” by Nikola Jurkovic
T
2:57Type 3 AudioNARRATOR
The META report on AI R&D automation within Anthropic is substantial evidence that timelines to superintelligence are now less than two years, even accounting for uncertainty about take-off modeling, the reliability of the report's conclusions, and other types of uncertainty.
T
3:13Type 3 AudioNARRATOR
Ryan Greenblatt gives a straightforward way to extrapolate META's findings in a tweet.
T
3:18Type 3 AudioNARRATOR
He claims that in AI 2027, the current level of speed-up by RSI happens around March 2027.
T
3:26Type 3 AudioNARRATOR
This means that real-world AI progress on AI R&D capabilities is faster than in AI 2027, an already extremely bullish scenario.
"Three Plans for 12-Month Timelines" by Nikola Jurkovic
T
2:57Type III AudioNARRATOR
The META report on AI R&D automation within Anthropic is substantial evidence that timelines to superintelligence are now less than two years, even accounting for uncertainty about take-off modeling, the reliability of the report's conclusions, and other types of uncertainty.
T
3:13Type III AudioNARRATOR
Ryan Greenblatt gives a straightforward way to extrapolate META's findings in a tweet.
T
3:18Type III AudioNARRATOR
He claims that in AI 2027, the current level of speed-up by RSI happens around March 2027.
T
3:26Type III AudioNARRATOR
This means that real-world AI progress on AI R&D capabilities is faster than in AI 2027, an already extremely bullish scenario.
Is Anthropic on the cusp of RSI? Inside the redacted METR report | Luisa Rodriguez
L
9:24Luisa RodriguezHOST
A lot of people I trust have changed their minds because of this report.
L
9:29Luisa RodriguezHOST
Ryan Greenblatt, who studies these risks for a living, says he was seriously wrong.
L
9:35Luisa RodriguezHOST
He'd expected much less speed up this year and now calls the report a big update towards superintelligence within one to two years.
L
9:44Luisa RodriguezHOST
Buck Schleggeris now thinks there's about a one in three chance AI companies build superintelligence within the next year.
Why Recursive Self-Improvement Hits a Wall | Ramez Naam
R
16:49Ramez NaamGUEST
There's a set of assumptions on how much coding uplift gets you, how much increase in the pace of AI.
R
16:59Ramez NaamGUEST
I think Ryan Greenblatt has talked about this, and the model is roughly like a square root.
R
17:04Ramez NaamGUEST
If you quadruple coding progress, number of lines of code per developer, let's say, you should get double the pace of AI progress.
R
17:13Ramez NaamGUEST
That's definitively not happening.
17 more episodes mention Ryan Greenblatt.
Create an account to see the whole feed, search across every transcript, and follow the entities you care about.