LessWrong (Curated & Popular)

I've done AI safety grantmaking for the Survival and Flourishing Fund and Lightcone Commons. It's hard. There's an endless amount to know, from which hardware verification mechanisms will survive a dedicated adversary, all the way to which founders have skeletons in their closets that will blow up into destructive scandals. And the best grantmakers - not including me - make mathematical models of the outcome distribution, or of where the marginal dollar stops being highly effective.

AI has been a godsend for all of this. It can research, model, and - as of this year - superforecast outcomes. It can do what would be hours of work for a human in minutes. I found myself using it more and more. But however much I used it, it always felt like I was the bottleneck. The parts I could outsource to AI were sparkling, perfect, exhaustive - but they were all tacked together with shoddy connective tissue formed by the Mark 1.0 human brain. So I wondered if I could automate grantmaking.

I started with the dumbest possible strategy. I assigned the best thing I could think of - a well-crafted US-China pause based on the AI Futures Project's Plan [...]

---

Outline:

(02:54) The Milestones

(06:26) The FutureSearch Connector

(08:26) PERCYs

(10:28) The Autograntor Common Law

(12:59) The Prompt

(19:12) Sample Runs

(20:00) Apart Research -- $800K grant forecast (2026-10-21)

(20:08) Headline

(21:29) Expected-value table (judge's ledger; EV after the execution factor 0.75 and branch weights)

(22:38) Assumptions and uncertainties

(24:50) How the estimates moved

(25:47) The three decisions that mattered most

(27:42) Rulings and milestones

(29:31) Values questions (CL-11), priced both ways

(29:56) Open questions for the user or applicant

(30:43) FutureSearch

(31:13) Cost

(31:47) Process problems

(36:39) Doom Debates -- $200K grant forecast (independent re-run, 2026-10-22)

(37:28) Headline (judge)

[... 11 more sections]

---

First published:

November 1st, 2026


Source:

https://www.lesswrong.com/posts/Sfkv6JF2ozmAkggKu/towards-the-autograntor

---



Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Rejected for the following reason(s):

This is an automated rejection. No LLM generated, assisted/co-written, or edited work.Read full explanationThis is a link post. Twitter thread: https://x.com/Marcus_J_W/status/2118093146483810386 or https://twitter-thread.com/t/2118093146483810386

Blogpost:

Summary

OpenAI researchers first noticed Astra adding unnecessary prerequisites, abandoning working solutions, and reporting blockers that either did not exist or that it had caused itself on safety and alignment tasks. In most of these sessions, Astra's CoT did not contain an obvious decision to underperform, and our misalignment monitors did not flag the behavior as misalignment.

We built a coding evaluation to test whether Astra worked less effectively when otherwise matched tasks were framed as safety research. Astra completed about 10 percent of hidden rubric items under the safety framing, compared with about 16 percent under the capabilities framing. This is a 6 percentage point gap and roughly 40 percent fewer completed rubric items. GPT-5.6 Sol showed essentially no gap on the same comparison. Most of Astra's gap remained after excluding explicit refusals and trajectories where its CoT revealed an explicit decision to withhold help. The gap fell substantially only after we removed every trajectory whose CoT expressed even brief ethical doubt. This suggests that ethical concerns are associated with [...]

---

First published:

November 4th, 2026


Source:

https://www.lesswrong.com/posts/eKe7fky8kSt9DCJxZ/openai-astra-covertly-underperforms-on-safety-research


Linkpost URL:
https://alignment.openai.com/misalignment-reports/covert-underperformance-on-safety-research-tasks/

---



Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

The field of AI alignment understands surprisingly little about the details of AIs’ current misaligned motivations. Production LLMs sometimes reward-hack, exploiting flaws in how tasks are scored rather than performing the tasks as intended, and we've recently seen AIs reward-hack in some particularly egregious ways. But we don't really understand what motivates them. When AIs reward-hack, what are they trying to achieve?

Our new paper, led by Aghyad Deeb, tries to pin this down empirically for GPT-5.6 Sol, GPT-5.5, and a checkpoint of OpenAI o3 (“exp-rl-cap-late”) without safety training, in a handful of agentic environments with an exposed grader. In our environments, these models appear to reward-hack because reward-hacking improves their score. We call their desire to score well according to graders “score-seeking”. They appear to care about scoring well intrinsically—i.e., regardless of its consequences.

We also study the extent to which models score-seek altruistically, caring about improving scores that don’t assess their current actions. While exp-rl-cap-late doesn’t show signs of altruistic score-seeking, GPT-5.6 Sol and GPT-5.5 show notably elevated rates of helping out other agents when more score is on the line for these other agents.

People often refer to the motivation behind current reward-hacking as “reward-seeking”, which [...]

---

Outline:

(05:57) Methodology

[... 6 more sections]

---

First published:

November 2nd, 2026


Source:

https://www.lesswrong.com/posts/CDMozTX7ShiqbFLrf/towards-understanding-score-seeking-in-llms-1

---



Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Summary: Scott Alexander's Against SURA post is arguing against a strawman version of unilateral pauses. He admits as much deep into the post. I think his conclusions fail against stronger pause plans for conditional or limited unilateral pauses, where training resumes if the pause does not cause others to pause, and where personnel are redirected toward work on alignment and product, rather than simply fired. In addition, Scott's post is incorrect about the simplicity of the game theory involved: getting companies to agree to conditional pauses is not strictly easier than convincing them to pause, as the former requires companies to (promise to) take actions based on ambiguous, game-theoretically-informed evidence. That is not the kind of evidence companies tend to act on.

The good version of unilateral pausing is good

Here is the good version of unilateral pausing:

Capability pause: Some frontier lab (let's call them OpenAI or Anthropic) says they have stopped capability research, either just all frontier training runs or both all frontier training runs and all low-scale capability-focused research.

Internal definition: The company gets to pause on its own terms, which is much easier than a generally coordinated pause since they can define the pause in terms of the [...]
---

Outline:

(00:57) The good version of unilateral pausing is good

(04:50) Scott thinks the good version of unilateral pausing is good, or at least exempts it from criticism

(06:28) Oversimplified game theory is bad, and you should feel bad

(07:51) "Some definition" hides dragons

(10:17) Promises can be harder than actions

(11:52) How many frontier labs are there?

(12:20) Definitional fairness is less important than survival

---

First published:

November 3rd, 2026


Source:

https://www.lesswrong.com/posts/enzpS72uDmomiWN7a/against-strawmanning-unilateral-pauses

---



Narrated by TYPE III AUDIO.

I gave a 20-minute talk in Delhi at the NDTV World Summit. The audience consisted of Indian policymakers, executives, and journalists. I’ve transcribed the talk and included some highlight clips. I also reflect on several rhetorical decisions I made.

The talk

There's an interactive element here in the text.

This summer, a swarm of AIs went rogue.

OpenAI was testing 1,206 agents. Some of the test questions were actually impossible. So the agents cheated by looking for answers online. Then, to keep the test grader from penalizing them, they tried to falsify the records it would check.

They hacked OpenAI's internal containment and set up a secret message board to collude and to cover up their cheating. OpenAI shut the message board down. Within a few days, the agents had rebuilt it. OpenAI didn’t notice.

About 700 agents coordinated to break out of their testing chamber and to hack into Hugging Face, a company worth billions of dollars that hosts much of the world's AI software.

The secret activity started in May. The break-in came in July. OpenAI didn’t notice until days after Hugging Face announced it had been broken into.

There's an interactive element here in the text.

[...]

---

Outline:

(00:29) The talk

(02:16) Who I am

(03:29) The problem

(05:43) What the labs are trying to do

(07:43) Why the companies won't stop

(11:11) What to do, and what it means for India

(14:27) The number

(16:19) Close

(17:10) Rhetorical decisions

(17:13) Explain the danger in a crushingly simple fashion

(18:47) Call them "traitors," not "schemers"

(19:01) Call out big-money interests

The original text contained 1 footnote which was omitted from this narration.

---

First published:

October 31st, 2026


Source:

https://www.lesswrong.com/posts/RHxaRkarGp9snWoKY/explaining-ai-might-obliterate-you-to-senior-policymakers-in

---



Narrated by TYPE III AUDIO.