prompt stringlengths 324 9.71k | target stringlengths 323 4.03k | meta_json stringlengths 147 171 | format stringclasses 1
value |
|---|---|---|---|
Below is a LessWrong post by Boaz Barak titled 'Six Thoughts on AI Safety'.
Post body (possibly truncated):
[Crossposted from windowsontheory]
The following statements seem to be both important for AI safety and are not widely agreed upon. These are my opinions, not those of my employer or colleagues. As is true for ... | I agree with not infinite and not being constant, but I do think the ratio for killing 90% is probably larger than 10x and plausibly much larger than 100x for some intermediate period of technological development. (Given realistic society adaptation and response.)
> As I mention, in an extreme case, governments might ... | {"comment_id": "chMfoTZid3L6HZnwn", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Boaz Barak", "post_id": "3jnziqCF3vA2NXAKp"} | dialogue |
Below is a LessWrong post by Sam Marks titled 'Towards Alignment Auditing as a Numbers-Go-Up Science'.
Post body (possibly truncated):
Thanks to Rowan Wang and Buck Shlegeris for feedback on a draft.
What is the job of an alignment auditing researcher? In this post, I propose the following potential answer: to build ... | This seems pretty reasonable (strong upvoted).
These aren't disagreements with the post, but here are some of my thoughts (most of which Sam already knows, but I'm sharing for third parties):
My largest concern is that the auditing test beds will end up being pretty disanalogous to the cases of misalignment I find mo... | {"comment_id": "mmouudcSjovCxe4cx", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Sam Marks", "post_id": "bGYQgBPEyHidnZCdE"} | dialogue |
Below is a LessWrong post by AdamGleave titled 'AI Safety in a World of Vulnerable Machine Learning Systems'.
Post body (possibly truncated):
Even the most advanced contemporary machine learning systems are vulnerable to adversarial attack. The safety community has often assumed adversarial robustness to be a problem ... | (Here's a possibly arcane remark. It's worth noting that I think *always correct* reward models are sufficient for high stakes alignment via runtime filtering (technically you just need to never give a very bad output decent reward). So, *always correct* reward models would be great if you could get them. Note that *al... | {"comment_id": "yphn7aBQGJ3AqJxCi", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "AdamGleave", "post_id": "ncsxcf8CkDveXBCrA"} | dialogue |
Below is a LessWrong post by Boaz Barak titled 'Machines of Faithful Obedience'.
Post body (possibly truncated):
[Crossposted on Windows On Theory]
Throughout history, technological and scientific advances have had both good and ill effects, but their overall impact has been overwhelmingly positive. Thanks to scien... | > I believe that AI (including AGI and ASI) can do the same and be a positive force for humanity. I also believe that it is possible to solve the “technical alignment” problem and build AIs that follow the words and intent of our instructions and report faithfully on their actions and observations.
>
> I will not defe... | {"comment_id": "4d8J2tinQMRa7Cadd", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Boaz Barak", "post_id": "faAX5Buxc7cdjkXQG"} | dialogue |
Below is a LessWrong post by Noosphere89 titled 'The case for multi-decade AI timelines [Linkpost]'.
Post body (possibly truncated):
So this post is an argument that multi-decade timelines are reasonable, and the key cruxes that Ege Erdil has with most AI safety people who believe in short timelines are due to the fol... | > I don’t see the trends that one would extrapolate in order to arrive at very short timelines on the order of a few years. The obvious trend extrapolations for AI’s economic impact give timelines to full remote work automation of around a decade, and I expect these trends to slow down by default.
Actually, I think t... | {"comment_id": "noxcihBEW3FGTFC43", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Noosphere89", "post_id": "xxxK9HTBNJvBY2RJL"} | dialogue |
Below is a LessWrong post by rosehadshar titled 'Three Types of Intelligence Explosion'.
Post body (possibly truncated):
Abstract
Once AI systems can design and build even more capable AI systems, we could see an intelligence explosion, where AI capabilities rapidly increase to well past human performance.
The class... | Some quick (and relatively minor) notes:
- I expect that full stack intelligence explosion could look more like "make the whole economy bigger using a bunch of AI labor" rather than specifically automating the chip production process. (That said, in practice I expect explicit focused automation of chip production to b... | {"comment_id": "RcKjDmZbtQcamqpuP", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "rosehadshar", "post_id": "PzbEpSGvwH3NnegDB"} | dialogue |
Below is a LessWrong post by aog titled 'Adversarial Robustness Could Help Prevent Catastrophic Misuse'.
Post body (possibly truncated):
There have been several discussions about the importance of adversarial robustness for scalable oversight. I’d like to point out that adversarial robustness is also important under a... | Agreed, I should have been more clear. Here I'm trying to argue about the question of whether people should work on research to mitigate misuse from the perspective of avoiding misuse through an API.
There is a separate question of reducing misuse concerns either via:
- Trying to avoid weights being stolen/leaking
- ... | {"comment_id": "CKNJnpAgpfjM3crbm", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "aog", "post_id": "timk6zHDTFdrHYLmu"} | dialogue |
Below is a LessWrong post by So8res titled 'Decision theory does not imply that we get to have nice things'.
Post body (possibly truncated):
(Note: I wrote this with editing help from Rob and Eliezer. Eliezer's responsible for a few of the paragraphs.)
A common confusion I see in the tiny fragment of the world that k... | IMO, this post makes several locally correct points, but overall fails to defeat the argument that misaligned AIs are somewhat likely to spend (at least) a tiny fraction of resources (e.g., between 1/million and 1/trillion) to satisfy the preferences of currently existing humans.
AFAICT, this is the main argument it w... | {"comment_id": "J8wTkp8CxoXsQNCfe", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "So8res", "post_id": "rP66bz34crvDudzcJ"} | dialogue |
Below is a LessWrong post by Raymond Douglas titled 'Gradual Disempowerment: Concrete Research Projects'.
Post body (possibly truncated):
This post benefitted greatly from comments, suggestions, and ongoing discussions with David Duvenaud, David Krueger, and Jan Kulveit. All errors are my own.
A few months ago, I and... | > That period where AIs are more capable than human, but human+AI is even more capable, seems like a particularly crucial window for doing useful things, so extending it is pretty valuable. In particular, both bringing forward augmented human capability, and also pushing back human redundance.
In the context of risks ... | {"comment_id": "YuiKa2dQnaeHphmCc", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Raymond Douglas", "post_id": "GAv4DRGyDHe2orvwB"} | dialogue |
Below is a LessWrong post by Zack_M_Davis titled 'And All the Shoggoths Merely Players'.
Post body (possibly truncated):
[Setting: a suburban house. The interior of the house takes up most of the stage; on the audience's right, we see a wall in cross-section, and a front porch. Simplicia enters stage left and rings th... | This is in the data constrained case right?
Maybe noise makes training worse because the model can't learn to just ignore it due to insufficient data? (E.g., making training more noisy means convergence/compute efficiency is lower.)
Also, does this decrease the size of the dataset by a factor of 5 in the uniform nois... | {"comment_id": "LhqnG3riSakbtKhjh", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Zack_M_Davis", "post_id": "8yCXeafJo67tYe5L4"} | dialogue |
Below is a LessWrong post by Sam Marks titled 'What’s up with LLMs representing XORs of arbitrary features?'.
Post body (possibly truncated):
Thanks to Clément Dumas, Nikola Jurković, Nora Belrose, Arthur Conmy, and Oam Patel for feedback.
In the comments of the post on Google Deepmind’s CCS challenges paper, I expre... | I think the key thing is that $a \land b$ and $a \lor b$ aren't *separately* linearly represented (or really even linearly represented in some strong sense). The model "represents" these by $a + b$ with different thresholds like this:
![](https://res.cloudinary.com/lesswrong-2-0/image/upload/f_auto,q_auto/v1/mirro... | {"comment_id": "sDj4jpPgsfJacApiJ", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Sam Marks", "post_id": "hjJXCn9GsskysDceS"} | dialogue |
Below is a LessWrong post by Max Harms titled 'Contra Collier on IABIED'.
Post body (possibly truncated):
Clara Collier recently reviewed If Anyone Builds It, Everyone Dies in Asterisk Magazine. I’ve been a reader of Asterisk since the beginning and had high hopes for her review. And perhaps it was those high hopes th... | > I can't actually think of that many cases of humans failing at aligning existing systems because the problem is too technically hard.
You're probably already tracking this, but the biggest cases of "alignment was actually pretty tricky" I'm aware are:
- Recent systems doing egregious reward hacking in some cases (i... | {"comment_id": "pLqBXfP9WdwLhWaTg", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Max Harms", "post_id": "JWH63Aed3TA2cTFMt"} | dialogue |
Below is a LessWrong post by joshc titled 'How might we safely pass the buck to AI?'.
Post body (possibly truncated):
My goal as an AI safety researcher is to put myself out of a job.
I don’t worry too much about how planet sized brains will shape galaxies in 100 years. That’s something for AI systems to figure out.
... | Note that the capability milestone forecasted in the linked short form is substantially weaker than the notion of transformative AI in the 2020 model. (It was defined as AI with an effect at least as large as the industrial revolution.)
I don't expect this adds many years, for me it adds like ~2 years to my median.
(... | {"comment_id": "z89CmZXfDT8j5Rtci", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "joshc", "post_id": "TTFsKxQThrqgWeXYJ"} | dialogue |
Below is a LessWrong post by Raemon titled 'The title is reasonable'.
Post body (possibly truncated):
Alt title: "I don't believe you that you actually disagree particularly with the core thesis of the book, if you pay attention to what it actually says."
I'm annoyed by various people who seem to be complaining ab... | Do you agree the feedback loops for capabilities are better right now?
> Sure, AI societies could go off the rails that hurts alignment R&D but not AI R&D; they could also go off the rails in a way that hurts AI R&D and not alignment R&D. Not sure why I should expect one rather than the other.
For this argument it's ... | {"comment_id": "KG9ikcvqabhJDM8YG", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Raemon", "post_id": "voEAJ9nFBAqau8pNN"} | dialogue |
Below is a LessWrong post by Akbir Khan titled 'Debating with More Persuasive LLMs Leads to More Truthful Answers'.
Post body (possibly truncated):
We've just completed a bunch of empirical work on LLM debate, and we're excited to share the results. If the title of this post is at all interesting to you, we recommend ... | Thanks for the response!
I think I agree with everything you said and I appreciate the level of thoughtfulness.
> Yeah we tried a bunch of other tasks early on, which we discuss in Appendix C.
Great! I appreciate the inclusion of negative results here.
> Of course this is not the same as human debaters who know the... | {"comment_id": "dEkYyCurSyejP42oj", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Akbir Khan", "post_id": "2ccpY2iBY57JNKdsP"} | dialogue |
Below is a LessWrong post by Zvi titled 'Joshua Achiam Public Statement Analysis'.
Post body (possibly truncated):
Joshua Achiam is the OpenAI Head of Mission Alignment
I start off this post with an apology for two related mistakes from last week.
The first is the easy correction: I incorrectly thought he was the he... | > Violently enforcing certain particularly important principles on non-signatories is entirely within the norm
True as stated, though I'm not aware of examples of this being enforced on non-signatories which are nuclear powers. This is just quantitatively riskier, not a notable change in norms.
And I agree this seem... | {"comment_id": "iBqnnv7SdgafxZ5TW", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Zvi", "post_id": "WavWheRLhxnofKHva"} | dialogue |
Below is a LessWrong post by evhub titled 'Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training'.
Post body (possibly truncated):
I'm not going to add a bunch of commentary here on top of what we've already put out, since we've put a lot of effort into the paper itself, and I'd mostly just reco... | Being able to squeeze useful work out of clearly misaligned models should probably be part of the portfolio. But, we might more minimally aim to just ensure that in cases where we're unsure if our model is scheming (aka deceptively aligned), it is at least directly safe (e.g. doesn't escape even if the work we get out ... | {"comment_id": "oNXbYMDYphHdePH2x", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "evhub", "post_id": "ZAsJv7xijKTfZkMtr"} | dialogue |
Below is a LessWrong post by Nina Panickssery titled 'Modulating sycophancy in an RLHF model via activation steering'.
Post body (possibly truncated):
Produced as part of the SERI ML Alignment Theory Scholars Program - Summer 2023 Cohort, under the mentorship of Evan Hubinger. Thanks to Alex Turner for his feedback an... | I'd guess that if you:
- Instructed human labelers to avoid sycophancy
- Gave human labelers examples of a few good and bad responses with respect to sycophancy
- Trained models on examples where sycophancy is plausibly/likely (e.g., pretrained models exhibit sycophancy a reasonable fraction of the time when generatin... | {"comment_id": "54hrsJT2msPP6CaWK", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Nina Panickssery", "post_id": "raoeNarFYCxxyKAop"} | dialogue |
Below is a LessWrong post by Bentham's Bulldog titled 'Eliezer Yudkowsky Is Frequently, Confidently, Egregiously Wrong'.
Post body (possibly truncated):
Introduction
Edit 8/27: I think the tone of this post was not ideal, changing much of it!
“After many years, I came to the conclusion that everything he says is fa... | > And as Schwarz points out, in the twin case, you'll get less utility by following FDT--you don't always want to be a FDTist.
I can't seem to find this in the linked blog post. (I see discussion of the twin case, but not a case where you get less utility from precommiting to follow FDT at the start of time.)
> I fi... | {"comment_id": "tXt2HCkutcMLwFWCA", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Bentham's Bulldog", "post_id": "TjyyngWFYvQWPpNNj"} | dialogue |
Below is a LessWrong post by jessicata titled 'On Eating the Sun'.
Post body (possibly truncated):
The Sun is the most nutritious thing that's reasonably close. It's only 8 light-minutes away, yet contains the vast majority of mass within 4 light-years of the Earth. The next-nearest star, Proxima Centauri, is about 4.... | The argument is (I assume):
- Once centuries have passed, you've already sent out huge amounts of space probes that roughly saturate reachable resources. (Because you can convert Proxima Centauri fully into probes within <20 years probably.)
- It doesn't take that much energy to pretty much fully saturate on probes. I... | {"comment_id": "7GXDo3mh94T8RCZmX", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "jessicata", "post_id": "6Fo8fjvpL7pwCTz3t"} | dialogue |
Below is a LessWrong post by evhub titled 'Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training'.
Post body (possibly truncated):
I'm not going to add a bunch of commentary here on top of what we've already put out, since we've put a lot of effort into the paper itself, and I'd mostly just reco... | > I think it's quite plausible you'd see clear features related to deception in our models without needing to have the backdoor triggers
Would you expect this to work better than just training a probe to identify lying/deception/scheming and seeing if it fires more on average? If so why?
As in, you train the probe "o... | {"comment_id": "ZcZ3enfahM3HuXbGL", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "evhub", "post_id": "ZAsJv7xijKTfZkMtr"} | dialogue |
Below is a LessWrong post by Roko titled '"AI Alignment" is a Dangerously Overloaded Term'.
Post body (possibly truncated):
Alignment as Aimability or as Goalcraft?
The Less Wrong and AI risk communities have obviously had a huge role in mainstreaming the concept of risks from artificial intelligence, but we have a s... | I think the overloading is actually worse than is discussed in this post, because people also sometimes use the term AI alignment to refer to "ensuring that AIs don't cause bad outcomes via whatever means".
For this problematic definition, it is possible to ensure "alignment" by using approaches like [AI control](http... | {"comment_id": "s85aAZtxAFgNRiBLB", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Roko", "post_id": "sy4whuaczvLsn9PNc"} | dialogue |
Below is a LessWrong post by Sam Marks titled 'What’s up with LLMs representing XORs of arbitrary features?'.
Post body (possibly truncated):
Thanks to Clément Dumas, Nikola Jurković, Nora Belrose, Arthur Conmy, and Oam Patel for feedback.
In the comments of the post on Google Deepmind’s CCS challenges paper, I expre... | Prediction: token level features or other extremely salient features are XOR'd with more things than less salient features. And if you find less salient things which are linearly represented, a bunch of this won't be XOR'd.
This solves the exponential blow up and should also make sense with your experimental results (... | {"comment_id": "XY4zv9yyXmBv3AbpD", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Sam Marks", "post_id": "hjJXCn9GsskysDceS"} | dialogue |
Below is a LessWrong post by Boaz Barak titled 'Mass surveillance, red lines, and a crazy weekend'.
Post body (possibly truncated):
[These are my own opinions, and not representing OpenAI. Cross-posted on windowsontheory.]
AI has so many applications, and AI companies have limited resources and attention span. Hen... | > What does "AI alignment" even mean at this point?
Responding at the object level:
- [Clarifying "AI Alignment"](https://www.lesswrong.com/posts/ZeE7EKHTFMBs8eMxn/clarifying-ai-alignment): "The “alignment problem” is the problem of building powerful AI systems that are aligned with their operators. This is significa... | {"comment_id": "KieNmHnDLvgrxA4tq", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Boaz Barak", "post_id": "zombjEubpz6pcPPHL"} | dialogue |
Below is a LessWrong post by otto.barten titled 'What Failure Looks Like is not an existential risk (and alignment is not the solution)'.
Post body (possibly truncated):
Introduction
Among those thinking that AI is an existential risk, there seems to be significant disagreement on what the main threat model is. Threa... | > Both threat models involve many AIs. In both threat models, there does not seem to be a deliberate AI takeover (e.g. caused by a resource conflict), either unipolar or multipolar. Rather, the danger is, according to this model, that things are ‘breaking’, rather than ‘taking’. The existential event would be accidenta... | {"comment_id": "mbh6u2sCJLK2hh7zg", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "otto.barten", "post_id": "tyHW6tEGzoErZXZ4x"} | dialogue |
Below is a LessWrong post by Bentham's Bulldog titled "Don't Eat Honey".
Post body (possibly truncated):
Crosspost from my blog.
(I think this is a pretty important article so I’d appreciate you sharing and restacking it—thanks!)
There are lots of people who say of themselves “I’m vegan except for honey.” This is a... | Hmm, I guess I think "something basically like hedonic utilitarianism, at least for downside" is pretty plausible.
Maybe a big difference is that I feel like I've generally updated away from putting much weight on moral intuitions / heuristics *except with respect to forbidding some actions because they violate norms,... | {"comment_id": "7kduyL2bbxzaSChZ4", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Bentham's Bulldog", "post_id": "CKEoFGuJ3C48p55ry"} | dialogue |
Below is a LessWrong post by Sam Marks titled 'Thoughts on open source AI'.
Post body (possibly truncated):
Epistemic status: I only ~50% endorse this, which is below my typical bar for posting something. I’m more bullish on “these are arguments which should be in the water supply and discussed” than “these arguments ... | > - Sadly, I’m not confident the answer is “yes,” and this is the main reason I only ~50% endorse this post. Two reasons I’m worried evaluators might fail:
> - [...]
> - The world might change in ways that enable new threat models after camelidAI is open-sourced. For example, suppose that camelidAI + GPT-SoTA isn... | {"comment_id": "JL5Z8Ahzbgj8aPqaB", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Sam Marks", "post_id": "WLYBy5Cus4oRFY3mu"} | dialogue |
Below is a LessWrong post by Raemon titled 'What percent of the sun would a Dyson Sphere cover?'.
Post body (possibly truncated):
I disagreed with a bunch of the implications of this comment, but I was curious about the specific question "Would a dyson sphere made out of the solar system necessarily cover (most of) th... | I don't think you can feasibly use the Hawking radiation of large black holes as an energy source in our universe (even if you are patient).
My understanding is that larger black holes decay over ~$10^{100}$ years. I did a botec a while ago and found that you maybe get 1 flop every $10^{30}$ years or so on average if ... | {"comment_id": "LfmaNEKzHcPuutDqm", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Raemon", "post_id": "EKhNDspRxLGcZAuJz"} | dialogue |
Below is a LessWrong post by Daniel Kokotajlo titled "Why Don't We Just... Shoggoth+Face+Paraphraser?".
Post body (possibly truncated):
Daniel KokotajloHere's a fairly concrete AGI safety proposal:
Default AGI design: Let's suppose we are starting with a pretrained LLM 'base model' and then we are going to do a to... | > Q: Why bother splitting shoggoth and face? Why not simply blind the reinforcement process to the internal CoT, like OpenAI claims to be doing with o1? You coulld still have a paraphraser too.
>
> A: Because if you use the same model weights to do both the internal reasoning and the external token outputs, there's mor... | {"comment_id": "7BkGjJcuGrKsBtY5b", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Daniel Kokotajlo", "post_id": "Tzdwetw55JNqFTkzK"} | dialogue |
Below is a LessWrong post by snewman titled '35 Thoughts About AGI and 1 About GPT-5'.
Post body (possibly truncated):
If this is GPT-5 in “Thinking” mode, I wonder what “Pro” mode looks like
Amidst the unrelenting tumult of AI news, it’s easy to lose track of the bigger picture. Here are some ideas that have been d... | In general, this post seems to make a bunch of claims that LLMs have specific large qualitative barriers relative to humans, but these claims seem mostly unsupported. The evidence as far as I can tell seems more consistent with LLMs being weaker in a bunch of specific quantitative ways which are improving over time. Fo... | {"comment_id": "QMiMFMAKqgnSweiQG", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "snewman", "post_id": "uAbbEz4p6tcsENaRz"} | dialogue |
Below is a LessWrong post by Daniel Kokotajlo titled 'AI 2027: What Superintelligence Looks Like'.
Post body (possibly truncated):
In 2021 I wrote what became my most popular blog post: What 2026 Looks Like. I intended to keep writing predictions all the way to AGI and beyond, but chickened out and just published up t... | > zero-shot WikiText103 perplexity and 5-shot MMLU
These are somewhat awkward benchmarks because they don't actually measure downstream usefulness at software engineering or AI research. In particular, these tasks might not measure improvements in RL which can have huge effects on usefulness and have seen fast algorit... | {"comment_id": "Ahd4tCfF2zFRsS9tp", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Daniel Kokotajlo", "post_id": "TpSFoqoG2M5MAAesg"} | dialogue |
Below is a LessWrong post by Sam Marks titled 'What’s up with LLMs representing XORs of arbitrary features?'.
Post body (possibly truncated):
Thanks to Clément Dumas, Nikola Jurković, Nora Belrose, Arthur Conmy, and Oam Patel for feedback.
In the comments of the post on Google Deepmind’s CCS challenges paper, I expre... | Yep I think I agree, I didn't understand the point you made about systematic anti-correlation originally.
If I understand correctly the issues is something like:
- There are 10 India related statements, exactly 5 of which are false and 5 of which are true.
- We do a random split of all the data, so if there is more t... | {"comment_id": "YuN4jXJpuyGEapK5C", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Sam Marks", "post_id": "hjJXCn9GsskysDceS"} | dialogue |
Below is a LessWrong post by Rubi J. Hudson titled 'The Case for Predictive Models'.
Post body (possibly truncated):
Thanks to Johannes Treutlein and Paul Colognese for feedback on this post.
Just over a year ago, the Conditioning Predictive Models paper was released. It laid out an argument and a plan for using powe... | For the proposed safety strategy (conditioning models to generate safety research based on alternative future worlds) to beat naive baselines (RLHF), you need:
- The CPM abstraction to hold extremely strongly in unlikely ways. E.g., models need to generalize basically like this.
- The advantage has to be coming from u... | {"comment_id": "faqmpBDqJ4GDLuwBm", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Rubi J. Hudson", "post_id": "RHDB3BdnvM233bnhG"} | dialogue |
Below is a LessWrong post by Lukas_Gloor titled 'We might be missing some key feature of AI takeoff; it\'ll probably seem like "we could\'ve seen this coming"'.
Post body (possibly truncated):
Predicting the future is hard, so it’s no surprise that we occasionally miss important developments.
However, several times r... | Relatedly, over time as capital demands increase, we might see huge projects which are collaborations between multiple countries.
I also think that investors could plausibly end up with more and more control over time if capital demands grow beyond what the largest tech companies can manage. (At least if these investo... | {"comment_id": "gZouQ84SsuPjWHZaj", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Lukas_Gloor", "post_id": "dLwo67p7zBuPsjG5t"} | dialogue |
Below is a LessWrong post by Nora Belrose titled 'Counting arguments provide no evidence for AI doom'.
Post body (possibly truncated):
Crossposted from the AI Optimists blog.
AI doom scenarios often suppose that future AIs will engage in scheming— planning to escape, gain power, and pursue ulterior motives, while dec... | > The key distinction in my view is whether the designers of the reward function intended for lies to be reinforced or not.
Hmm, I don't think the intention is the key thing (at least with how I use the word and how I think Joe uses the word), I think the key thing is whether the reinforcement/reward process actively... | {"comment_id": "LwxxsgrvGhqae8XD2", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Nora Belrose", "post_id": "YsFZF3K9tuzbfrLxo"} | dialogue |
Below is a LessWrong post by Buck titled 'An epistemic advantage of working as a moderate'.
Post body (possibly truncated):
[epistemic status: the points I make are IMO real and important, but there are also various counterpoints; I'm not settled on an overall opinion here, and the categories I draw are probably kind ... | > Carlsmith's Multiple Stage Fallacy risk estimate of 5% that involved only an 80% chance anyone would even try to build agentic AI?
This is false as stated. The report says:
The corresponding footnote ... | {"comment_id": "PcEzyXPibLyzLhPDk", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Buck", "post_id": "9MaTnw5sWeQrggYBG"} | dialogue |
Below is a LessWrong post by Raemon titled 'The title is reasonable'.
Post body (possibly truncated):
Alt title: "I don't believe you that you actually disagree particularly with the core thesis of the book, if you pay attention to what it actually says."
I'm annoyed by various people who seem to be complaining ab... | > I... really don't know what Scott expected a story that featured actual superintelligence to look like. I think the authors bent over backwards giving us one of the least-sci-fi stories you could possibly tell that includes superintelligence doing anything at all, without resorting to "superintelligence just won't ev... | {"comment_id": "QqCkofRQ7kRXQqJBf", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Raemon", "post_id": "voEAJ9nFBAqau8pNN"} | dialogue |
Below is a LessWrong post by AdamGleave titled 'AI Safety in a World of Vulnerable Machine Learning Systems'.
Post body (possibly truncated):
Even the most advanced contemporary machine learning systems are vulnerable to adversarial attack. The safety community has often assumed adversarial robustness to be a problem ... | > Suppose we condition on RLHF failing. At a high level, failures split into: (a) human labelers rewarded the wrong thing (e.g. fooling humans); (b) the reward model failed to predict human labelers judgement and rewarded the wrong thing (e.g. reward hacking); (c) RL produced a policy that is capable enough to be dange... | {"comment_id": "xAyYsRJjwcd7mb5hN", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "AdamGleave", "post_id": "ncsxcf8CkDveXBCrA"} | dialogue |
Below is a LessWrong post by Lucius Bushnaq titled 'Interpretability: Integrated Gradients is a decent attribution method'.
Post body (possibly truncated):
A short post laying out our reasoning for using integrated gradients as attribution method. It is intended as a stand-alone post based on our LIB papers [1] [2]. T... | [Not very confident, but just saying my current view.]
I'm pretty skeptical about integrated gradients.
As far as why, I don't think we should care about the derivative at the baseline (zero or the mean).
As far as the axioms, I think I get off the train on "Completeness" which doesn't seem like a property we need/w... | {"comment_id": "KipdKgFELqk4dprJj", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Lucius Bushnaq", "post_id": "Rv6ba3CMhZGZzNH7x"} | dialogue |
Below is a LessWrong post by TurnTrout titled 'Dreams of AI alignment: The danger of suggestive names'.
Post body (possibly truncated):
Let's not forget the old, well-read post: Dreams of AI Design. In that essay, Eliezer correctly points out errors in imputing meaning to nonsense by using suggestive names to describe... | Are you claiming that future powerful AIs won't be well described as pursuing goals (aka being goal-directed)? This is the read I get from the the "dragon" analogy you mention, but this can't possibly be right because AI agents are already obviously well described as pursuing goals (perhaps rather stupidly). TBC the go... | {"comment_id": "XqxzGktE9WuxdkLAF", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "TurnTrout", "post_id": "yxWbbe9XcgLFCrwiL"} | dialogue |
Below is a LessWrong post by snewman titled "AI 2027 is a Bet Against Amdahl's Law".
Post body (possibly truncated):
EDIT: I've written a followup post, summarizing and responding to the key themes raised in the comments.
AI 2027 lies at a Pareto frontier – it contains the best researched argument for short timeli... | ("Has to" is maybe a bit strong, I think I probably should have said "will probably end up needing to be better competitive with the best human experts at basically everything (other than vision) and better at more central AI R&D given the realistic capability profile". I think I generally expect full automation to hit... | {"comment_id": "TSRjhsmctLSRiKTtg", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "snewman", "post_id": "bfHDoWLnBH9xR3YAK"} | dialogue |
Below is a LessWrong post by Akbir Khan titled 'Debating with More Persuasive LLMs Leads to More Truthful Answers'.
Post body (possibly truncated):
We've just completed a bunch of empirical work on LLM debate, and we're excited to share the results. If the title of this post is at all interesting to you, we recommend ... | I disagreed due to a combination of 2, 3, and 4. (Where 5 feeds into 2 and 3).
For 4, the upside is just that the title is less long and confusingly caveated.
Norms around titles seem ok to me given issues with space.
Do you have issues with our recent paper title "[AI Control: Improving Safety Despite Intentional Su... | {"comment_id": "zrzteFik4s8r7keRq", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Akbir Khan", "post_id": "2ccpY2iBY57JNKdsP"} | dialogue |
Below is a LessWrong post by snewman titled 'What Indicators Should We Watch to Disambiguate AGI Timelines?'.
Post body (possibly truncated):
(Cross-post from https://amistrongeryet.substack.com/p/are-we-on-the-brink-of-agi, lightly edited for LessWrong. The original has a lengthier introduction and a bit more explana... | Suppose we get an AI system which can (at least) automate away the vast majority of the job of a research engineer at an AI company (e.g. OpenAI). Let's say this results in an increase in productivity among OpenAI capabilities researchers at least equivalent to the productivity you'd get as if the human employees opera... | {"comment_id": "sJSGDKTKcZgNwHvZP", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "snewman", "post_id": "auGYErf5QqiTihTsJ"} | dialogue |
Below is a LessWrong post by Joe Carlsmith titled 'Can we safely automate alignment research?'.
Post body (possibly truncated):
(This is the fifth essay in a series that I’m calling “How do we solve the alignment problem?”. I’m hoping that the individual essays can be read fairly well on their own, but see this introd... | You might hope for elicitation efficiency, as in, you heavily RL the model to produce useful considerations and hope that your optimization is good enough that it covers everything well enough.
Or, two lower bars you might hope for:
- It brings up considerations that it "knows" about. (By "knows" I mean relatively de... | {"comment_id": "gLtFb534fMkgFPadQ", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Joe Carlsmith", "post_id": "nJcuj4rtuefeTRFHp"} | dialogue |
Below is a LessWrong post by Rafael Harth titled '≤10-year Timelines Remain Unlikely Despite DeepSeek and o3'.
Post body (possibly truncated):
[Thanks to Steven Byrnes for feedback and the idea for section §3.1. Also thanks to Justis from the LW feedback team.]
Remember this?
Or this?
The images are from WaitButWhy... | I mean, I don't think AI R&D is a particularly hard field persay, but I do think it involves lots of tricky stuff and isn't much easier than automating some other plausibly-important-to-takeover field (e.g., robotics). (I could imagine that the AIs have a harder time automating philosophy even if they were trying to wo... | {"comment_id": "aawHBx4eTPXJGtEud", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Rafael Harth", "post_id": "gsj3TWdcBxwkm9eNt"} | dialogue |
Below is a LessWrong post by Zach Stein-Perlman titled 'OpenAI: Preparedness framework'.
Post body (possibly truncated):
OpenAI released a beta version of their responsible scaling policy (though they don't call it that). See summary page, full doc, OpenAI twitter thread, and Jan Leike twitter thread [edit: and Zvi co... | [Minor correction to notes]
> There's some commitment that the Board will be in the loop. To my knowledge, this is the first commitment by a frontier lab to give their Board specific information or specific power besides removing-the-CEO.
The Anthropic RSP has a bunch of stuff along these lines:
> Follow an "Update ... | {"comment_id": "6cenBuMzDkCY3DNzh", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Zach Stein-Perlman", "post_id": "oPbiQfRotHYuC3wfE"} | dialogue |
Below is a LessWrong post by Thane Ruthenis titled '"Humanity vs. AGI" Will Never Look Like "Humanity vs. AGI" to Humanity'.
Post body (possibly truncated):
When discussing AGI Risk, people often talk about it in terms of a war between humanity and an AGI. Comparisons between the amounts of resources at both sides' di... | A thing I always feel like I'm missing in your stories of how the future goes is: "if it is obvious that the AIs are exerting substantial influence and acquiring money/power, why don't people train competitor AIs which don't take a cut?"
A key difference between AIs and immigrants is that it might be relatively easy t... | {"comment_id": "6NrRRpPEJGrt4wqpZ", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Thane Ruthenis", "post_id": "xSJMj3Hw3D7DPy5fJ"} | dialogue |
Below is a LessWrong post by Akbir Khan titled 'Debating with More Persuasive LLMs Leads to More Truthful Answers'.
Post body (possibly truncated):
We've just completed a bunch of empirical work on LLM debate, and we're excited to share the results. If the title of this post is at all interesting to you, we recommend ... | I like this paper, but I think the abstract is somewhat overstated. In particular, instead of:
> We find that debate consistently helps both non-expert models and humans answer questions,
I wish this was something more like:
> On the QuALITY dataset and in the case where debators are given more knowledge than otherw... | {"comment_id": "kDymqfSgMevYBFDjc", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Akbir Khan", "post_id": "2ccpY2iBY57JNKdsP"} | dialogue |
Below is a LessWrong post by So8res titled 'Ability to solve long-horizon tasks correlates with wanting things in the behaviorist sense'.
Post body (possibly truncated):
Status: Vague, sorry. The point seems almost tautological to me, and yet also seems like the correct answer to the people going around saying “LLMs t... | > Anyhow I think this is mostly just a misunderstanding of Nate and my position. It doesn't contradict anything we've said.
I think it contradicts things Nate says in this post directly. I don't know if it contradicts things you've said.
To clarify, I'm commenting on the following chain:
First Nate said:
> This obs... | {"comment_id": "aJTJxFBjn8FStEPP3", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "So8res", "post_id": "AWoZBzxdm4DoGgiSj"} | dialogue |
Below is a LessWrong post by So8res titled 'Apocalypse insurance, and the hardline libertarian take on AI risk'.
Post body (possibly truncated):
Short version: In a saner world, AI labs would have to purchase some sort of "apocalypse insurance", with premiums dependent on their behavior in ways that make reckless beha... | Why is this a problem? People who are interested in the long run can buy these property rights while people who don't care can sell them.
If AIs respect these property rights[^ensure] but systematically care more about the long run future, then so be it. I expect that in practice some people will explicitly care about... | {"comment_id": "xB6TpkDsJscrB8tZy", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "So8res", "post_id": "mSeesg7i4d9scWAet"} | dialogue |
Below is a LessWrong post by Gretta Duleba titled 'MIRI 2024 Communications Strategy'.
Post body (possibly truncated):
As we explained in our MIRI 2024 Mission and Strategy update, MIRI has pivoted to prioritize policy, communications, and technical governance research over technical alignment research. This follow-up... | > building misaligned smarter-than-human systems will kill everyone, including their children [...] if they come to understand this central truth.
I'd like to once again reiterate that the arguments for misaligned AIs killing literally all humans (if they succeed in takeover) are quite weak and probably literally all ... | {"comment_id": "xBYimQtgASti5tgWv", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Gretta Duleba", "post_id": "tKk37BFkMzchtZThx"} | dialogue |
Below is a LessWrong post by Joe Rogero titled 'LLMs are badly misaligned'.
Post body (possibly truncated):
This post is part of the sequence Against Muddling Through.
A core objection to If Anyone Builds It, Everyone Dies seems to run through the intuition that modern LLMs are some flavor of partially aligned, or at... | To clarify my current perspective: I don't think current LLMs (released as of Oct 2025) plot against us or intentionally sabotage work to achieve their longer-run aims (for the vast, vast majority of typical usage; you can construct situations where they do this). I'd guess this isn't controversial. Correspondingly, I ... | {"comment_id": "AuKnZmsjLz2arHbgw", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Joe Rogero", "post_id": "QdEHS9TmBBKirz7Ky"} | dialogue |
Below is a LessWrong post by Robert_AIZI titled "Comments on Anthropic's Scaling Monosemanticity".
Post body (possibly truncated):
These are some of my notes from reading Anthropic's latest research report, Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet.
TL;DR
In roughly descending o... | It's seems plausible to me that a 70b model stores ~6 billion bits of memorized information.
Naively, you might think this requires around 500M features. (supposing that each "feature" represents 12 bits which is probably a bit optimistic)
I don't think SAEs will actually work at this level of sparsity though, so this... | {"comment_id": "pA3LLfAte3L4cZxhG", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Robert_AIZI", "post_id": "zzmhsKx5dBpChKhry"} | dialogue |
Below is a LessWrong post by Oliver Sourbut titled 'Deceptive Alignment and Homuncularity'.
Post body (possibly truncated):
NB this dialogue occurred at the very end of 2023, and for various reasons is only being published ~a year later! Keep this in mind while reading.
Key topics covered
consistent NN-internal dece... | > I was pretty convinced that pre-training (or light post-training) was unlikely to produce a coherent deceptively aligned agent, [as we discuss in that paper](https://www.lesswrong.com/s/n3utvGrgC2SGi9xQX/p/qoHwKgLFfPcEuwaba). [...] (maybe ~1% likely).
When I look up your views as of 2 years ago, [it appears that yo... | {"comment_id": "YjmusHaDEZtuAcagw", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Oliver Sourbut", "post_id": "9htmQx5wiePqTtZuL"} | dialogue |
Below is a LessWrong post by Nikola Jurkovic titled 'Thoughts on extrapolating time horizons'.
Post body (possibly truncated):
(written for a Twitter audience)
Has AI progress slowed down? I’ll write some personal takes and predictions in this post.
The main metric I look at is METR’s time horizon, which measures th... | > More concretely, my median is that AI research will be automated by the end of 2028, and AI will be better than humans at >95% of current intellectual labor by the end of 2029.
It seems like you think AI research will be (fully?) automated once AIs are at 80% reliability on 1 month long benchmarkable / easy-to-check... | {"comment_id": "cdvowJr7Hth8XAhDD", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Nikola Jurkovic", "post_id": "GAJbegsvnd85hX3eS"} | dialogue |
Below is a LessWrong post by Nina Panickssery titled 'Modulating sycophancy in an RLHF model via activation steering'.
Post body (possibly truncated):
Produced as part of the SERI ML Alignment Theory Scholars Program - Summer 2023 Cohort, under the mentorship of Evan Hubinger. Thanks to Alex Turner for his feedback an... | **Edit**: This comment now seems kinda silly as you basically addressed this in your comment and I missed it, feel free to ignore.
> Also, as I predicted, the benefits stack with those of finetuning and in-context learning.
For the task of removing sycophancy this isn't clearly true right? As you note in the linked p... | {"comment_id": "AAKJGnmjuywoMnqbx", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Nina Panickssery", "post_id": "raoeNarFYCxxyKAop"} | dialogue |
Below is a LessWrong post by Buck titled 'A freshman year during the AI midgame: my approach to the next year'.
Post body (possibly truncated):
I recently spent some time reflecting on my career and my life, for a few reasons:
It was my 29th birthday, an occasion which felt like a particularly natural time to think t... | I still like this post overall, but various things have changed that interestingly affect the content of the post:
- We'd now be just starting the 3rd year of university in Buck's analogy. Does this seem right? I guess maybe. It feels a bit late to me. (Maybe I feel more like a 2nd year.)
- Redwood is doing less blue-... | {"comment_id": "gg9nr57JPxZztxiNX", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Buck", "post_id": "mfFn46AEiBL5EzaXr"} | dialogue |
Below is a LessWrong post by Boaz Barak titled 'Machines of Faithful Obedience'.
Post body (possibly truncated):
[Crossposted on Windows On Theory]
Throughout history, technological and scientific advances have had both good and ill effects, but their overall impact has been overwhelmingly positive. Thanks to scien... | I think if the attitude in AI was "there can't be any even slightly plausible routes to misalignment related catastrophe" and this was consistently upheld in a reasonable way, that would address my concerns. (So, e.g. by the time we're deploying AIs which could cause huge problems if they were conspiring against us the... | {"comment_id": "wdm5uwTHSKMivhNbS", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Boaz Barak", "post_id": "faAX5Buxc7cdjkXQG"} | dialogue |
Below is a LessWrong post by Joe Rogero titled 'We won’t get AIs smart enough to solve alignment but too dumb to rebel'.
Post body (possibly truncated):
This post is part of the sequence Against Muddling Through.
I often hear it proposed that AIs which are “aligned enough” to their developers may help solve alignment... | > To clarify, by "align systems..." did you mean the same thing I do, full-blown value alignment / human CEV?
No, I mean "make AI robustly pursue the intended aims in practice when deferring to them on doing safety research and managing the situation" which IMO requires something much weaker, though for sufficiently p... | {"comment_id": "ZbSaE8m86R6AJskKo", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Joe Rogero", "post_id": "8buEtNxCScYpjzgW8"} | dialogue |
Below is a LessWrong post by snewman titled 'Interpreting the METR Time Horizons Post'.
Post body (possibly truncated):
[EDIT: initial publication did not have the link to the original post, and was missing the footnotes. Sorry about that! Fixed now.]
[This has been lightly edited from the original post, eliminating ... | To be clear, I agree it provides evidence against very aggressive timelines (if I had 2027 medians I would have updated to longer), I was disagreeing with "the study doesn’t say much about AGI, except to". I think the study does provide a bunch of evidence about when AGI might come! (And it seems you agree.) I edited m... | {"comment_id": "qy8BGNeJjwKTcJ7my", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "snewman", "post_id": "fRiqwFPiaasKxtJuZ"} | dialogue |
Below is a LessWrong post by snewman titled "AI 2027 is a Bet Against Amdahl's Law".
Post body (possibly truncated):
EDIT: I've written a followup post, summarizing and responding to the key themes raised in the comments.
AI 2027 lies at a Pareto frontier – it contains the best researched argument for short timeli... | > Please let me know if you suspect I've over-interpreted that validation.
Slightly? My view is more like:
- For AIs to be superhuman AI researchers, they probably need to match humans at most underlying/fundamental cognitive tasks, including reasonably sample efficient learning. (Or at least learning which is compet... | {"comment_id": "x2gqgjzajDrNGCkKn", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "snewman", "post_id": "bfHDoWLnBH9xR3YAK"} | dialogue |
Below is a LessWrong post by Rob Bensinger titled 'The Problem'.
Post body (possibly truncated):
This is a new introduction to AI as an extinction threat, previously posted to the MIRI website in February alongside a summary. It was written independently of Eliezer and Nate's forthcoming book, If Anyone Builds It, Eve... | > By what mechanism? This feels like 'we get a pause' or 'there's a wall'.
I just meant that takeoff isn't that fast so we have like >0.5-1 year at a point where AIs are at least very helpful for safety work (if reasonably elicited) which feels plausible to me. The duration of "AIs could fully automate safety (includi... | {"comment_id": "atovvQQ94AaCuqp7H", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Rob Bensinger", "post_id": "kgb58RL88YChkkBNf"} | dialogue |
Below is a LessWrong post by Raemon titled 'What percent of the sun would a Dyson Sphere cover?'.
Post body (possibly truncated):
I disagreed with a bunch of the implications of this comment, but I was curious about the specific question "Would a dyson sphere made out of the solar system necessarily cover (most of) th... | I think the size might have to be pretty precise to get this right (I think decay duration is cubic in mass), so they'd probably need to be engineered to have a particular size. (E.g. add mass to a small black hole until it hits the right size.)
But, yeah, with this constraint, I think it can maybe work. (I don't know... | {"comment_id": "THWAwauecRthcKzQh", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Raemon", "post_id": "EKhNDspRxLGcZAuJz"} | dialogue |
Below is a LessWrong post by Steven Byrnes titled 'Foom & Doom 1: “Brain in a box in a basement”'.
Post body (possibly truncated):
1.1 Series summary and Table of Contents
This is a two-post series on AI “foom” (this post) and “doom” (next post).
A decade or two ago, it was pretty common to discuss “foom & doom” sce... | > On the foom side, Paul Christiano brings up Eliezer Yudkowsky’s past expectation that ASI “would likely emerge from a small group rather than a large industry” as a failed prediction [here \[disagreement 12\]](https://www.lesswrong.com/posts/CoZhXrhpQxpy9xw9y/where-i-agree-and-disagree-with-eliezer#Disagreements) and... | {"comment_id": "jy6Lfs7rawgPFEonv", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Steven Byrnes", "post_id": "yew6zFWAKG4AGs3Wk"} | dialogue |
Below is a LessWrong post by snewman titled '35 Thoughts About AGI and 1 About GPT-5'.
Post body (possibly truncated):
If this is GPT-5 in “Thinking” mode, I wonder what “Pro” mode looks like
Amidst the unrelenting tumult of AI news, it’s easy to lose track of the bigger picture. Here are some ideas that have been d... | > But if this were true, you’d think they’d be able to handle ARC-AGI puzzles (see the example image just above)
In a footnote you note that models do well on ARC-AGI-1, but I think you're description of the situation is misleading:
* AIs trained on the training set of ARC-AGI (and given a bunch of compute) can *be... | {"comment_id": "b9F9J3vxMrnHN574p", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "snewman", "post_id": "uAbbEz4p6tcsENaRz"} | dialogue |
Below is a LessWrong post by Beth Barnes titled 'Bounty: Diverse hard tasks for LLM agents'.
Post body (possibly truncated):
Update 3/14/2024: This post is out of date. For current information on the task bounty, see our Task Development Guide.
Summary
METR (formerly ARC Evals) is looking for (1) ideas, (2) detai... | I think these tasks would be in scope as long as these tasks can be done fully digitally and can be done relatively easily in text. (And it's possible to setup the task in a docker container etc etc.)
Quoting the desiderata from the post:
> Plays to strengths of LLM agents: ideally, most of the tasks can be completed... | {"comment_id": "Bt25ve9SsGh59esK4", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Beth Barnes", "post_id": "gAkCCaBBHD4gcwxmv"} | dialogue |
Below is a LessWrong post by Richard_Ngo titled 'Towards a scale-free theory of intelligent agency'.
Post body (possibly truncated):
I recently left OpenAI to pursue independent research. I’m working on a number of different research directions, but the most fundamental is my pursuit of a scale-free theory of intellig... | > However, this strongly limits the space of possible aggregated agents. Imagine two EUMs, Alice and Bob, whose utilities are each linear in how much cake they have. Suppose they’re trying to form a new EUM whose utility function is a weighted average of their utility functions. Then they’d only have three options:
>
>... | {"comment_id": "a7mR7p2ZbHpgNY6Pq", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Richard_Ngo", "post_id": "5tYTKX4pNpiG4vzYg"} | dialogue |
Below is a LessWrong post by Nicholas Kross titled '[SEE NEW EDITS] No, *You* Need to Write Clearer'.
Post body (possibly truncated):
This post is aimed solely at people in AI alignment/safety.
EDIT 3 October 2023: This post did not even mention, let alone account for, how somebody should post half-baked/imperfect/ha... | I can't tell if this post is trying to discuss communicating about anything related to AI or alignment or is trying to more specifically discuss communication aimed at general audiences. I'll assume it's discussing arbitrary communication on AI or alignment.
I feel like this post doesn't engage sufficiently with the c... | {"comment_id": "seMfCZxnLmByWaJPo", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Nicholas Kross", "post_id": "mLubC65xXekk5tkug"} | dialogue |
Below is a LessWrong post by Andrew_Critch titled 'Acausal normalcy'.
Post body (possibly truncated):
This post is also available on the EA Forum.
Summary: Having thought a bunch about acausal trade — and proven some theorems relevant to its feasibility — I believe there do not exist powerful information hazards abou... | > Simulations are not the most efficient way for A and B to reach their agreement
Are you claiming that the marginal returns to simulation are *never* worth the costs? I'm skeptical. I think it's quite likely that some number of acausal trade simulations are run even if that isn't where most of the information comes f... | {"comment_id": "5Z2azFypoeaXJ6TLE", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Andrew_Critch", "post_id": "3RSq3bfnzuL3sp46J"} | dialogue |
Below is a LessWrong post by snewman titled 'What Indicators Should We Watch to Disambiguate AGI Timelines?'.
Post body (possibly truncated):
(Cross-post from https://amistrongeryet.substack.com/p/are-we-on-the-brink-of-agi, lightly edited for LessWrong. The original has a lengthier introduction and a bit more explana... | Sure, but I think that at the relevant point, you'll probably be spending at least 5x more on experiments than on inference and potentially a much larger larger ratio if heavy test time compute usage isn't important. I was just trying to argue that the naive inference cost isn't that crazy.
Notably, if you give each r... | {"comment_id": "WYxNp47njNXKShbL8", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "snewman", "post_id": "auGYErf5QqiTihTsJ"} | dialogue |
Below is a LessWrong post by Nora Belrose titled 'Counting arguments provide no evidence for AI doom'.
Post body (possibly truncated):
Crossposted from the AI Optimists blog.
AI doom scenarios often suppose that future AIs will engage in scheming— planning to escape, gain power, and pursue ulterior motives, while dec... | > Since there are “more” possible schemers than non-schemers, the argument goes, we should expect training to produce schemers most of the time. In Carlsmith’s words:
It's important to note that the exact counting argument you quote isn't one that Carlsmith endorses, just one that he is explaning. And in fact Carlsmit... | {"comment_id": "QNBftRXT4gCtH4q4g", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Nora Belrose", "post_id": "YsFZF3K9tuzbfrLxo"} | dialogue |
Below is a LessWrong post by Nora Belrose titled 'Counting arguments provide no evidence for AI doom'.
Post body (possibly truncated):
Crossposted from the AI Optimists blog.
AI doom scenarios often suppose that future AIs will engage in scheming— planning to escape, gain power, and pursue ulterior motives, while dec... | I found the explanation at the point where you introduce $b$ confusing.
Here's a revised version of the text there that would have been less confusing to me (assuming I haven't made any errors):
> - Complexity of simplest deceptive objective: $l + b$ where $l$ is the number of bits needed to select the part of the ob... | {"comment_id": "7H7qsxH9vgFnJMq3Q", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Nora Belrose", "post_id": "YsFZF3K9tuzbfrLxo"} | dialogue |
Below is a LessWrong post by Buck titled 'An epistemic advantage of working as a moderate'.
Post body (possibly truncated):
[epistemic status: the points I make are IMO real and important, but there are also various counterpoints; I'm not settled on an overall opinion here, and the categories I draw are probably kind ... | I think williawa's characterization of how people reacted to bioanchors basically matches my experience and I'm skeptical of the claim that OpenPhil was very politically controlling of EA with respect to timelines.
And, I agree with the claim that that Eliezer often implies people interpreted bioanchors in some way th... | {"comment_id": "Zg7RNyF9iQrm3qQs5", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Buck", "post_id": "9MaTnw5sWeQrggYBG"} | dialogue |
Below is a LessWrong post by Nora Belrose titled "Deconstructing Bostrom's Classic Argument for AI Doom".
Post body (possibly truncated):
I had a pretty great discussion with social psychologist and philosopher Lance Bush recently about the orthogonality thesis, which ended up turning into a broader analysis of Nick B... | As far as the orthogonality thesis, relevant context is:
- The [arbital page which defines it more precisely](https://arbital.com/p/orthogonality/): "The Orthogonality Thesis asserts that there can exist arbitrarily intelligent agents pursuing any kind of goal."
- [Yudkowsky's tweet](https://twitter.com/ESYudkowsky/st... | {"comment_id": "8bNnQizLHJrgGCTYz", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Nora Belrose", "post_id": "RbynKk3evb6RiLryL"} | dialogue |
Below is a LessWrong post by Zvi titled 'AI #107: The Misplaced Hype Machine'.
Post body (possibly truncated):
The most hyped event of the week, by far, was the Manus Marketing Madness. Manus wasn’t entirely hype, but there was very little there there in that Claude wrapper.
Whereas here in America, OpenAI dropped an... | The exact text is:
> Dario: 6 months ago I I made this prediction that, you know, in in 6 months 90% of code would be written by by by AI models. Some people think that prediction is wrong, but within Anthropic and within a number of companies that we work with, that is absolutely true.
>
> Marc: Um now 90 you know s... | {"comment_id": "TfPJhJf4hYkF5rXY8", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Zvi", "post_id": "XFGTJz9vGwjJADeFB"} | dialogue |
Below is a LessWrong post by HoldenKarnofsky titled 'How might we align transformative AI if it’s developed very soon?'.
Post body (possibly truncated):
This post is part of my AI strategy nearcasting series: trying to answer key strategic questions about transformative AI, under the assumption that key events will ha... |
> We can't be confident enough that it won't happen to safely rely on that assumption.
I'm not sure what motivation for [worst-case reasoning](https://www.lesswrong.com/posts/yTvBSFrXhZfL8vr5a/worst-case-thinking-in-ai-alignment) you're thinking about here. Maybe just that there are many disjunctive ways things can g... | {"comment_id": "mJJ3ejDuW9zw8N38t", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "HoldenKarnofsky", "post_id": "rCJQAkPTEypGjSJ8X"} | dialogue |
Below is a LessWrong post by Joe Carlsmith titled 'The stakes of AI moral status'.
Post body (possibly truncated):
Podcast version (read by the author) here, or search for "Joe Carlsmith Audio" on your podcast app.
1. Introduction
Currently, most people treat AIs like tools. We act like AIs don’t matter in themselve... | > It seems very puzzling to me that almost no one is working on increasing AI and/or human philosophical competence in these ways, or even publicly expressing the worry that AIs and/or humans collectively might not be competent enough to solve important philosophical problems that will arise during and after the AI tra... | {"comment_id": "urnQ7RgbdX6bdyDsd", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Joe Carlsmith", "post_id": "tr6hxia3T8kYqrKm5"} | dialogue |
Below is a LessWrong post by titotal titled '"Diamondoid bacteria" nanobots: deadly threat or dead-end? A nanotech investigation '.
Post body (possibly truncated):
Confidence level: I’m a computational physicist working on nanoscale simulations, so I have some understanding of most of the things discussed here, but I ... | > By the way, where's this number coming from? (10^30 FLOP) You keep repeating it.
Extremely rough and slightly conservatively small ball park number for how many FLOP will be used to create powerful AIs. The idea being that this will represent roughly how many FLOP could plausibly be available at the time.
GPT-4 is ... | {"comment_id": "kkrZokwPMGpJpmR8v", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "titotal", "post_id": "bc8Ssx5ys6zqu3eq9"} | dialogue |
Below is a LessWrong post by Fabien Roger titled 'An issue with training schemers with supervised fine-tuning'.
Post body (possibly truncated):
This is a rough write-up of ideas by Ryan Greenblatt. The core ideas are his and mistakes are mine. Thanks to him, Buck Shlegeris and to Aghyad Deeb for feedback on the draft ... | > this failure mode is dangerous because of scheming AI and I say it's dangerous because the policy is OOD
I would say that it is dangerous in the case where is is both OOD enough that the AI can discriminate and the AI is scheming.
Neither alone would present a serious (i.e. catastrophic) risk in the imitation conte... | {"comment_id": "vnxG9dHmt6S7tervb", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Fabien Roger", "post_id": "mXYdYh6L9odTJZDSm"} | dialogue |
Below is a LessWrong post by Bruce W. Lee titled 'Distillation Robustifies Unlearning'.
Post body (possibly truncated):
Current “unlearning” methods only suppress capabilities instead of truly unlearning the capabilities. But if you distill an unlearned model into a randomly initialized model, the resulting network is... | I wrote up [some thoughts on how distillation can be used for AI safety here](https://www.lesswrong.com/posts/8KKujApx4g7FBm6hE/ai-safety-techniques-leveraging-distillation). The most relevant section of this post is [this section about using distillation for more precise capability control](https://www.lesswrong.com/p... | {"comment_id": "tyw3wAzaWdKGwCdtC", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Bruce W. Lee", "post_id": "anX4QrNjhJqGFvrBr"} | dialogue |
Below is a LessWrong post by Adam Scholl titled 'Pay Risk Evaluators in Cash, Not Equity'.
Post body (possibly truncated):
Personally, I suspect the alignment problem is hard. But even if it turns out to be easy, survival may still require getting at least the absolute basics right; currently, I think we're mostly fai... | I raised a similar proposal to various people a while ago.
The strongest objection I'm aware of is something like:
> You actually want evaluators to have as much skin in the game as other employees so that when they take actions that might shut the company down or notably reduce the value of equity, this is a costly... | {"comment_id": "q5vfGDSkZupyXWySQ", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Adam Scholl", "post_id": "sMBjsfNdezWFy6Dz5"} | dialogue |
Below is a LessWrong post by Adam Karvonen titled 'Using an LLM perplexity filter to detect weight exfiltration'.
Post body (possibly truncated):
A recent area of focus has been securing AI model weights. If the weights are located in a data center and an adversary wants to obtain model weights, the weights have to le... | Sadly, I don't think this proposal works. More precisely, testing if data is low perplexity is strictly worse than compressing the data against a small language model and then actually sending out a small payload.
---
> Encoding this much data in a way that appears as natural language would be extremely challenging. ... | {"comment_id": "iGoPYLBc8JgHGBzSD", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Adam Karvonen", "post_id": "aWZEDw6oxR6Wk5hru"} | dialogue |
Below is a LessWrong post by YafahEdelman titled 'Report on Frontier Model Training'.
Post body (possibly truncated):
Understanding what drives the rising capabilities of AI is important for those who work to forecast, regulate, or ensure the safety of AI. Regulations on the export of powerful GPUs need to be informed... | My cached state is that the A/H100 vs 4090 price gap is mostly price discrimination rather than a large difference in the actual manufacturing cost.
I think price discrimination is very common in computing hardware and nvidia happens to have a quite powerful monopoly right now for various reasons.
Note that [4090s te... | {"comment_id": "hwndp2MyvN8u9aRCe", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "YafahEdelman", "post_id": "nXcHe7t4rqHMjhzau"} | dialogue |
Below is a LessWrong post by habryka titled 'AI Timelines'.
Post body (possibly truncated):
Introduction
How many years will pass before transformative AI is built? Three people who have thought about this question a lot are Ajeya Cotra from Open Philanthropy, Daniel Kokotajlo from OpenAI and Ege Erdil from Epoch. De... | My sense is that this post holds up pretty well. Most of the considerations under discussion still appear live and important including: in-context learning, robustness, whether jank AI R&D accelerating AIs can quickly move to more general and broader systems, and general skepticism of crazy conclusions.
At the time of... | {"comment_id": "qLnpLE7EAxGjER7iv", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "habryka", "post_id": "K2D45BNxnZjdpSX2j"} | dialogue |
Below is a LessWrong post by Logan Riggs titled "When AI 10x's AI R&D, What Do We Do?".
Post body (possibly truncated):
Note: below is a hypothetical future written in strong terms and does not track my actual probabilities.
Throughout 2025, a huge amount of compute is spent on producing data in verifiable tasks, suc... | By top 3 priority, I mean "among the top 3 most prioritized *cyber attacks* of that year". Precisely, I'm discussing robustness against OC5 as defined in the RAND report linked above:
> OC5 Top-priority operations by the top cyber-capable institutions
>
> Operations roughly less capable than or comparable to 1,000 i... | {"comment_id": "ntmrZjFvqyn43m3kY", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Logan Riggs", "post_id": "SvvYCH6JrLDT8iauA"} | dialogue |
Below is a LessWrong post by Oleg Trott titled 'Fix simple mistakes in ARC-AGI, etc.'.
Post body (possibly truncated):
ARC-AGI is a diverse artificial dataset that aims to test general intelligence. It's sort of like an IQ test that's played out on rectangular grids.
Last month, @ryan_greenblatt proposed an approach ... | Seems like a reasonable idea. To implement this, I'd have to look more carefully at exactly what types of mistakes GPT-4o makes to calibrate what should/shouldn't be dithered. (Additional programs are cheap, but you can easily get a combinatorial explosion with this sort of thing.)
(I'm not currently working on ARC-AG... | {"comment_id": "NLLrmDe6yJv8JN9TL", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Oleg Trott", "post_id": "jEEWe42fcJWdbCZo9"} | dialogue |
Below is a LessWrong post by habryka titled 'Using Dangerous AI, But Safely?'.
Post body (possibly truncated):
Rob Miles has released a new video, this time on control!
Youtube Transcript (Very rough)
0:00
hi so this channel is about AI safety a lot of people when they hear AI safety
0:05
the first thing that com... | To the extent that complexity makes succesfully red-teaming our approaches too difficult, then we could try to mostly depend on simple approaches and only use more complex approaches only if we are confident that at least they won't can't make the situation worse.
My general view is that complex methods are only produ... | {"comment_id": "Xm5u6RGkeCgye4j2n", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "habryka", "post_id": "TBLztZivgMbpYCPvt"} | dialogue |
Below is a LessWrong post by evhub titled 'RSPs are pauses done right'.
Post body (possibly truncated):
COI: I am a research scientist at Anthropic, where I work on model organisms of misalignment; I was also involved in the drafting process for Anthropic’s RSP. Prior to joining Anthropic, I was a Research Fellow at M... | This post seems mostly reasonable in retrospect, except that it doesn't specifically note that it seems [unlikely that voluntary RSP commitments would result in AI companies unilaterally pausing until they were able to achieve broadly reasonable levels of safety](https://www.lesswrong.com/posts/neTbrpBziAsTH5Bn7/ai-com... | {"comment_id": "y2EPa9beey68Le36E", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "evhub", "post_id": "mcnWZBnbeDz7KKtjJ"} | dialogue |
Below is a LessWrong post by Jesse Hoogland titled 'Announcing Timaeus'.
Post body (possibly truncated):
Timaeus is a new AI safety research organization dedicated to making fundamental breakthroughs in technical AI alignment using deep ideas from mathematics and the sciences. Currently, we are working on singular lea... | > The strongest critique of developmental interpretability we know is the following: while it is established that phase transitions exist in neural network training, it is not yet clear how common they are, and whether they make a good target for alignment.
Is it established that phase transitions exist in the trainin... | {"comment_id": "G2cGSSQ6qkgzwPApH", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Jesse Hoogland", "post_id": "nN7bHuHZYaWv9RDJL"} | dialogue |
Below is a LessWrong post by snewman titled '35 Thoughts About AGI and 1 About GPT-5'.
Post body (possibly truncated):
If this is GPT-5 in “Thinking” mode, I wonder what “Pro” mode looks like
Amidst the unrelenting tumult of AI news, it’s easy to lose track of the bigger picture. Here are some ideas that have been d... | > Is sample-efficient learning a singularly important step on the path to AGI?
Almost definitionally, learning as efficiently as top humans would suffice for AGI. (You could just train the AI on way more data/compute and it would be superhuman.)
AIs will probably reach milestones like full automation of AI R&D before... | {"comment_id": "G8CwrsqbvJzv8EDiL", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "snewman", "post_id": "uAbbEz4p6tcsENaRz"} | dialogue |
Below is a LessWrong post by Eliezer Yudkowsky titled 'Universal Basic Income and Poverty'.
Post body (possibly truncated):
(Crossposted from Twitter)
I'm skeptical that Universal Basic Income can get rid of grinding poverty, since somehow humanity's 100-fold productivity increase (since the days of agriculture) didn... | > you should also be wary of 'minor clarity improvements' suggested by ChatGPT/Claude. I find a lot of them make prose worse, especially if you apply most of them so the gestalt becomes ChatGPTese.
I agree with this, my editing process is to get the new output, view the diff and then copy in changes one-by-one if they... | {"comment_id": "qfxBLnb3p3XruDnZd", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Eliezer Yudkowsky", "post_id": "fPvssZk3AoDzXwfwJ"} | dialogue |
Below is a LessWrong post by Gretta Duleba titled 'MIRI 2024 Communications Strategy'.
Post body (possibly truncated):
As we explained in our MIRI 2024 Mission and Strategy update, MIRI has pivoted to prioritize policy, communications, and technical governance research over technical alignment research. This follow-up... | > That phrase sounds like the Terminator movies to me; it sounds like plucky humans could still band together to overthrow their robot overlords. I want to convey a total loss of control.
Yeah, seems like a reasonable concern.
FWIW, I also do think that it is reasonably likely that we'll see conflict between human fa... | {"comment_id": "qFbX8c997hQenqWvb", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Gretta Duleba", "post_id": "tKk37BFkMzchtZThx"} | dialogue |
Below is a LessWrong post by Alex Mallen titled 'Balancing Label Quantity and Quality for Scalable Elicitation'.
Post body (possibly truncated):
ArXiv paper.
Thanks to Nora Belrose, Buck Shlegeris, Jan Hendrik Kirchner, and Ansh Radhakrishnan for guidance throughout the project.
Scalable oversight studies methods of... | This paper seems like a great first stab at starting to answer [very important questions about scalable oversight sample efficiency](https://www.lesswrong.com/posts/6AT4vhYzww56CR6cm/scalable-oversight-as-a-quantitative-rather-than-qualitative) (at least for non-scheming models). Better understanding of sample effiicie... | {"comment_id": "HH4rRoZdBAKmrcRdj", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Alex Mallen", "post_id": "iA87ke4so2eqiJfsi"} | dialogue |
Below is a LessWrong post by Joe Carlsmith titled 'Video and transcript of talk on "Can goodness compete?"'.
Post body (possibly truncated):
(This is the video and transcript of a public talk I gave at Mox in San Francisco in July 2025, on long-term equilibria post-AGI. It’s a longer version of the talk I gave at this... | Here are two other potentially serious failures of strategy stealing[^comp]:
- [Vacuum decay](https://www.lesswrong.com/posts/zteMisMhEjwhZbWez/vacuum-decay-expert-survey-results) might allow for agents who control small amounts of resources to destory everything if there isn't relevant policing/prevention everywhere.... | {"comment_id": "pKjkGdodNFXRedcyh", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Joe Carlsmith", "post_id": "evYne4Xx7L9J96BHW"} | dialogue |
Below is a LessWrong post by Sam Bowman titled 'The Checklist: What Succeeding at AI Safety Will Involve '.
Post body (possibly truncated):
Crossposted by habryka with Sam's permission. Expect lower probability for Sam to respond to comments here than if he had posted it (he said he'll be traveling a bunch in the comi... | My proposal would roughly be that the US government (in collaboration with allies etc) enforces no one building AI which are qualitatively smarter than humans and this should be the default plan.
(This might be doable without government support via coordination between multiple labs, but I basically doubt it.)
Their ... | {"comment_id": "eANGMjfpxNcrnBJvm", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Sam Bowman", "post_id": "mGCcZnr4WjGjqzX5s"} | dialogue |
Below is a LessWrong post by snewman titled 'Interpreting the METR Time Horizons Post'.
Post body (possibly truncated):
[EDIT: initial publication did not have the link to the original post, and was missing the footnotes. Sorry about that! Fixed now.]
[This has been lightly edited from the original post, eliminating ... | > In actuality, the study doesn’t say much about AGI, except to provide evidence against the most aggressive forecasts.
This feels quite wrong to me. Surely if AIs were completing 1 month long self contained software engineering tasks (e.g. what a smart intern might do in the first month) that would be a big update to... | {"comment_id": "JieqtgAspBYFh3CCf", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "snewman", "post_id": "fRiqwFPiaasKxtJuZ"} | dialogue |
Below is a LessWrong post by Nora Belrose titled 'Counting arguments provide no evidence for AI doom'.
Post body (possibly truncated):
Crossposted from the AI Optimists blog.
AI doom scenarios often suppose that future AIs will engage in scheming— planning to escape, gain power, and pursue ulterior motives, while dec... | > I definitely thought you were making a counting argument over function space
I've argued multiple times that Evan was not *intending* to make a counting argument in function space:
- In discussion with Alex Turner (TurnTrout) when commenting on an earlier draft of this post.
- In discussion with Quintin after shari... | {"comment_id": "TWaBGZLmNgDFZaBZw", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Nora Belrose", "post_id": "YsFZF3K9tuzbfrLxo"} | dialogue |
Below is a LessWrong post by David Duvenaud titled 'Sabotage Evaluations for Frontier Models'.
Post body (possibly truncated):
This is a linkpost for a new research paper from the Alignment Evaluations team at Anthropic and other researchers, introducing a new suite of evaluations of models' abilities to undermine mea... | > My guess would be that they're not saying that well-designed control evaluations become untrustworthy
It's a bit messy because we have some ability to check whether we should be able to evaluate things.
So, there are really three relevant "failure" states for well done control:
1. We can't find countermeasures suc... | {"comment_id": "jtw34wJQhJ5haJDNi", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "David Duvenaud", "post_id": "nnvn6kaBLajiicH8e"} | dialogue |
Below is a LessWrong post by Zack_M_Davis titled 'My Interview With Cade Metz on His Reporting About Slate Star Codex'.
Post body (possibly truncated):
On 16 March 2024, I sat down to chat with New York Times technology reporter Cade Metz! In part of our conversation, transcribed below, we discussed his February 2021 ... | > The benefits are transparency about who is influencing society
In this particular case, I don't really see any transparency benefits. If it was the case that there was important public information attached to Scott's full name, then this argument would make sense to me.
(E.g. if Scott Alexander was actually Mark Zu... | {"comment_id": "5R9Kf2HZEkkM6x3wS", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Zack_M_Davis", "post_id": "oYnwTuxySiaZYDrur"} | dialogue |
Below is a LessWrong post by Raemon titled 'Anthropic, and taking "technical philosophy" more seriously'.
Post body (possibly truncated):
So, I have a lot of complaints about Anthropic, and about how EA / AI safety people often relate to Anthropic (i.e. treating the company as more trustworthy/good than makes sense).... | Yep, just the obvious. (I'd say "much less bought in" than "isn't bought in", but whatever.)
I don't really have dots I'm trying to connect here, but this feels more central to me than what you discuss. Like, I think "alignment might be really, really hard" (which you focus on) is less of the crux than "is misalignmen... | {"comment_id": "frp73opG8oTogtxBd", "format": "dialogue", "parent_kind": "external_post_body", "post_author": "Raemon", "post_id": "7uTPrqZ3xQntwQgYz"} | dialogue |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.