garrison

• garrison 11 Feb 2025 17:04 UTC

  1 point

  0

  in reply to: rvnnt’s comment on: Why Did Elon Musk Just Offer to Buy Con­trol of OpenAI for $100 Billion?

  Fixed, thanks!

• garrison 16 Jan 2025 17:49 UTC

  1 point

  0

  in reply to: _liminaldrift’s comment on: Is AI Hit­ting a Wall or Mov­ing Faster Than Ever?

  I think that is a problem for the industry, but probably not an insurmountable barrier the way some commentators make it out to be.

  1. o-series of models may be able to produce new high quality training data

  2. sufficiently good reasoning approaches + existing base models + scaffolding may be sufficient to get you to automating ML research

  One other thought is that there’s probably an upper limit on how good an LLM can get even with unlimited high quality data and I’d guess that models would asymptotically approach it for a while. Based on the reporting around GPT-5 and other next-gen models, I’d guess that the issue is lack of data rather than approaching some fundamental limit.

• garrison 23 Dec 2024 20:10 UTC

  4 points

  0

  in reply to: Edy Nastase’s comment on: We are in a New Paradigm of AI Progress—OpenAI’s o3 model makes huge gains on the tough­est AI bench­marks in the world

  It was all my twitter feed was talking about, but I think it’s been really under-discussed in mainstream press.

  RE Knoop’s comment, here are some relevant grafs from the ARC announcement blog post:

  To adapt to novelty, you need two things. First, you need knowledge – a set of reusable functions or programs to draw upon. LLMs have more than enough of that. Second, you need the ability to recombine these functions into a brand new program when facing a new task – a program that models the task at hand. Program synthesis. LLMs have long lacked this feature. The o series of models fixes that.

  For now, we can only speculate about the exact specifics of how o3 works. But o3′s core mechanism appears to be natural language program search and execution within token space – at test time, the model searches over the space of possible Chains of Thought (CoTs) describing the steps required to solve the task, in a fashion perhaps not too dissimilar to AlphaZero-style Monte-Carlo tree search. In the case of o3, the search is presumably guided by some kind of evaluator model. To note, Demis Hassabis hinted back in a June 2023 interview that DeepMind had been researching this very idea – this line of work has been a long time coming.

  More in the ARC post.

  My rough understanding is that it’s like a meta-CoT strategy, evaluating multiple different approaches.

• garrison 28 Nov 2024 19:21 UTC

  9 points

  1

  in reply to: Adrian Garcia’s comment on: China Hawks are Man­u­fac­tur­ing an AI Arms Race

  This is what Hsu just said about it: “3. I could be described as a China hawk in that I’ve been pointing to a US-China competition as unavoidable for over a decade. But I think I have more realistic views about what is happening in PRC than most China hawks. I also try to focus on simple descriptive analysis rather than getting distracted by normative midwit stuff.”

  https://​​x.com/​​hsu_steve/​​status/​​1861970671527510378

• garrison 28 Nov 2024 19:21 UTC

  11 points

  2

  in reply to: gwern’s comment on: China Hawks are Man­u­fac­tur­ing an AI Arms Race

  Steve Hsu clarified some things on my thread about this discussion: https://​​x.com/​​hsu_steve/​​status/​​1861970671527510378

  “Clarifications:

  1. The mafia tendencies (careerist groups working together out of self-interest and not to advance science itself) are present in the West as well these days. In fact the term was first used in this way by Italian academics.

  2. They’re not against big breakthroughs in PRC, esp. obvious ones. The bureaucracy bases promotions, raises, etc. on metrics like publications in top journals, cititations, … However there are very obvious wins that they will go after in a coordinated way—including AI, semiconductors, new energy tech, etc.

  3. I could be described as a China hawk in that I’ve been pointing to a US-China competition as unavoidable for over a decade. But I think I have more realistic views about what is happening in PRC than most China hawks. I also try to focus on simple descriptive analysis rather than getting distracted by normative midwit stuff.

  4. There is coordinated planning btw govt and industry in PRC to stay at the frontier in AI/​AGI/​ASI. They are less susceptible to “visionaries” (ie grifters) so you’ll find fewer doomers or singularitarians, etc. Certainly not in the top govt positions. The quiet confidence I mentioned extends to AI, not just semiconductors and other key technologies.”

• garrison 21 Nov 2024 23:52 UTC

  1 point

  0

  in reply to: Raemon’s comment on: China Hawks are Man­u­fac­tur­ing an AI Arms Race

  Gotcha, well I’m on it!

• garrison 21 Nov 2024 23:51 UTC

  4 points

  1

  in reply to: Logan Zoellner’s comment on: China Hawks are Man­u­fac­tur­ing an AI Arms Race

  Interesting, do you have a link for that?

  US companies are racing toward AGI but the USG isn’t. As someone else mentioned, Dylan Patel from Semianalysis does not think China is scale-pilled.

• garrison 21 Nov 2024 4:18 UTC

  11 points

  2

  in reply to: Raemon’s comment on: China Hawks are Man­u­fac­tur­ing an AI Arms Race

  As mentioned in another reply, I’m planning to do a lot more research and interviews on this topic, especially with people who are more hawkish on China. I also think it’s important that unsupported claims with large stakes get timely pushback, which is in tension with the type of information gathering you’re recommending (which is also really important, TBC!).

• garrison 21 Nov 2024 4:15 UTC

  4 points

  1

  in reply to: Jacob Pfau’s comment on: China Hawks are Man­u­fac­tur­ing an AI Arms Race

  Claiming that China as a country is racing toward AGI != Chinese AI companies aren’t fast following US AI companies, which are explicitly trying to build AGI. This is a big distinction!

• garrison 21 Nov 2024 4:12 UTC

  10 points

  6

  in reply to: Seth Herd’s comment on: China Hawks are Man­u­fac­tur­ing an AI Arms Race

  Hey Seth, appreciate the detailed engagement. I don’t think the 2017 report is the best way to understand what China’s intentions are WRT to AI, but there was nothing in the report to support Helberg’s claim to Reuters. I also cite multiple other sources discussing more recent developments (with the caveat in the piece that they should be taken with a grain of salt). I think the fact that this commission was not able to find evidence for the “China is racing to AGI” claim is actually pretty convincing evidence in itself. I’m very interested in better understanding China’s intentions here and plan to deep dive into it over the next few months, but I didn’t want to wait until I could exhaustively search for the evidence that the report should have offered while an extremely dangerous and unsupported narrative takes off.

  I also really don’t get the error pushback. These really were less technical errors than basic factual errors and incoherent statements. They speak to a sloppiness that should affect how seriously the report should be taken. I’m not one to gatekeep ai expertise, but idt it’s too much to expect a congressional commission with a top recommendation to commence in a militaristic AI arms race to have SOMEONE read a draft who knows that chatgpt-3 isn’t a thing.

• garrison 14 Nov 2024 16:16 UTC

  1 point

  0

  in reply to: Vladimir_Nesov’s comment on: Is Deep Learn­ing Ac­tu­ally Hit­ting a Wall? Eval­u­at­ing Ilya Sutskever’s Re­cent Claims

  Thanks for these!

• garrison 13 Nov 2024 22:10 UTC

  6 points

  0

  in reply to: Noosphere89’s comment on: Is Deep Learn­ing Ac­tu­ally Hit­ting a Wall? Eval­u­at­ing Ilya Sutskever’s Re­cent Claims

  I think this is a misunderstanding of the piece and how journalists typically paraphrase things. The reporters wrote that Ilya told them that results from scaling up pre-training have plateaued. So he probably said something to that effect, but for readability and word-count reasons, they paraphrased it.

  If a reported story from a credible outlet says something like X told us that Y, then the reporters are sourcing claim Y to X, whether or not they include a direct quote.

  The plateau claim also jives with The Information story about OpenAI, as well as a few other similar claims made by people in industry.

  Ilya probably spoke to the reporter(s) for at least a few min, so the quotes you see are a tiny fraction of everything he said.

• garrison 17 Jul 2024 15:40 UTC

  9 points

  0

  in reply to: Ruby’s comment on: Against Aschen­bren­ner: How ‘Si­tu­a­tional Aware­ness’ con­structs a nar­ra­tive that un­der­mines safety and threat­ens humanity

  FWIW I was also confused by this usage of sic, bc I’ve only ever seen it as indicating the error was in the original quote. Quotes seem sufficient to indicate you’re quoting the original piece. I use single quotes when I’m not quoting a specific person, but introducing a hypothetical perspective.

• garrison 28 Mar 2024 22:38 UTC

  2 points

  0

  in reply to: trevor’s comment on: My In­ter­view With Cade Metz on His Re­port­ing About Slate Star Codex

  I only skimmed the NYT piece about China and ai talent, but didn’t see evidence of what you said (dishonestly angle shooting the AI safety scene).

• garrison 5 Mar 2024 18:13 UTC

  1 point

  0

  in reply to: Nathan Helm-Burger’s comment on: Claude Doesn’t Want to Die

  The fey thing stuck out to me too. I’ll guess ChatGPT?

  I agree that it’s hard to disentangle the author/​character thing. I’m really curious for what the base model would say about its situation (especially without the upstream prompt “You are a language model developed by...”).

• garrison 14 Feb 2024 3:18 UTC

  2 points

  0

  in reply to: brambleboy’s comment on: My cover story in Ja­cobin on AI cap­i­tal­ism and the x-risk debates

  Thank you so much! I haven’t gotten any serious negative feedback from lefties for the EA stuff so far, though an e/​acc on Twitter mentioned it haha

• garrison 12 Feb 2024 21:56 UTC

  2 points

  1

  in reply to: Logan Zoellner’s comment on: Sam Alt­man’s Chip Am­bi­tions Un­der­cut OpenAI’s Safety Strategy

  Maybe I wasn’t clear enough in the writing, but I make basically the same point about the desirability of a slow takeoff in the piece.

• garrison 10 Feb 2024 19:56 UTC

  13 points

  20

  on: OpenAI wants to raise 5-7 trillion

  This approach appears to directly contradict Altman’s blogpost from less than a year ago arguing for short timelines + slow takeoff because of less compute overhang. I wrote more on this here.

• garrison 28 Oct 2022 15:50 UTC

  1 point

  0

  in reply to: the gears to ascension’s comment on: Pod­cast: The Left and Effec­tive Altru­ism with Habiba Islam

  I’m exploring adding transcripts, and would do this one retroactively.

  Good to know RE YouTube. I haven’t uploaded there before (it’s outside of the RSS workflow and I’m not sure how much it would expand reach), but seeing comments like this is helpful info.