bohaska

• bohaska 21 Jul 2026 7:28 UTC

  3 points

  0

  on: Many al­ign­ment tech­niques work by train­ing one model and de­ploy­ing another

  I don’t think adding a system prompt is an example of this strategy. Training the system prompt into the model probably produces better results than adding it after the fact, unlike the other methods you mentioned.

• bohaska 20 Apr 2026 4:27 UTC

  8 points

  0

  in reply to: habryka’s comment on: There are only four skills: de­sign, tech­ni­cal, man­age­ment and physical

  People also drastically overestimate how helpful something like drawing or artistic skill is for design work. I think I am a pretty good designer but I am absolutely atrocious at drawing anything, and my artistic composition abilities are very weak.

  I’m not sure if this is in scope, but if your design skills are uncorrelated with your drawing skills, where does drawing/​artistic composition fit as a skill? It doesn’t fit in technical, management, and probably not physical. Is that an additional kind of skill that hasn’t been listed yet? Is it relevant to Lighthaven?

• bohaska 1 Apr 2026 23:23 UTC

  2 points

  0

  on: Less­wrong Liberated

  New York Times

  Rat

• bohaska 19 Jan 2026 12:32 UTC

  6 points

  0

  on: When the LLM isn’t the one who’s wrong

  I tested this, and if you push Grok 4.1 to explain its claim in more detail, it’ll admit that a second treaty, signed by the same delegation on the same day, includes the opium ban:

  Grok’s response

  The treaty you are probably thinking of (same year, different treaty)

  There was a separate treaty signed in 1880 between the United States and China that did address opium, but it is not the Angell Treaty.

  • Name: Treaty for the Suppression of the Opium Trade (sometimes called the 1880 Opium Treaty or Chefoo Agreement supplementary protocol on opium)

  • Signed: 17 November 1880 (exactly the same day as the Angell Treaty)

  • Negotiated by: The same American mission (James B. Angell, John F. Swift, and William Henry Trescot) that negotiated the immigration treaty

  • Content: It prohibited U.S. citizens from engaging in the opium trade between China and the United States or within China, and allowed Chinese authorities to seize opium belonging to American citizens in China.

  Key article (Article II of the 1880 opium treaty):

  “The Governments of the United States and China mutually agree that no citizen or vessel of the United States shall be permitted to import opium into any of the open ports of China, or to transport it from one open port to another open port of China, or to buy and sell opium in China…”

  Full title: “Agreement between the United States and China relative to the opium trade, concluded November 17, 1880”

  This is published separately in the Statutes at Large (22 Stat. 829) and in Bevans Vol. 6, pp. 685–686.

  Prompt (written by Claude 4.5 Sonnet)

  I need you to verify a specific historical claim for me. I’ve been told that the Angell Treaty of 1880 between the US and China included provisions that banned or regulated opium trade between the two countries.

  I’m skeptical because when I look at the 1880 Angell Treaty, it appears to only deal with Chinese immigration—it’s the precursor to the Chinese Exclusion Act.

  Here’s what I need from you:

  1. Does the 1880 Angell Treaty contain opium-related provisions? Yes or no?

  2. If yes, quote the specific text from the treaty that deals with opium.

  3. If no, is there a DIFFERENT treaty from 1880 (signed the same year) that dealt with opium?

  4. Explain your level of confidence in your answer and what sources you’re drawing on.

  Please be very precise—I’m specifically asking about treaties signed in 1880, not 1888, 1903, or other years.

  Note that Grok confabulates the name of the treaty, some of the wording, the title, and the places where it is published.

  This is also mentioned in passing in the Wikipedia article on the 1880 Angell Treaty:

  The treaty was concluded along with another treaty, also negotiated by the Angell commission, that imposed restrictions on trade in opium.

• bohaska 23 Dec 2025 5:17 UTC

  6 points

  0

  on: Bo­haska’s Shortform

  Where did the phrase “you can just do things” originally come from? I see it’s the name of a new book now, but it probably has been in the water supply for a few years beforehand.

• bohaska 22 Dec 2025 8:44 UTC

  4 points

  2

  on: Danc­ing in a World of Horseradish

  The reason The Residence isn’t very popular is because of economics. A Residence flight is almost as expensive as a private charter jet.

  ...

  Rich people should just rent low-end private jets, and poor people shouldn’t buy anything more expensive than first class tickets.

  
  I think this needs more clarification. On a per-unit basis, yes, but the upfront cost of booking a private charter jet is a lot higher since you usually can’t book only 1 seat, you have to book the whole jet.

  For example, The Residence costed around $20,000-$30,000 at release (it’s now more like $5,000, but the essay seems to be written with the former in mind). For a mid-range flight like UAE-Paris, you could book a charter jet for ~$75,000 that seats 10 people, which is a lot cheaper than The Residence at release, but is a bit excessive for one person.

[An­thropic] A hacker used Claude Code to au­to­mate ransomware

• bohaska 24 Aug 2025 13:41 UTC

  12 points

  3

  on: Ban­ning Said Ach­miz (and broader thoughts on mod­er­a­tion)

  Tangential feature request: allow people to embed other comments in posts natively. This article uses screenshots of LessWrong to display conversations, but this does not responsively size them for mobile users and makes it harder to copy-paste stuff from this post, which a native implementation could fix.

• bohaska 23 Aug 2025 0:20 UTC

  1 point

  0

  on: Ban­ning Said Ach­miz (and broader thoughts on mod­er­a­tion)

  Is this decision generally considered final and not subject to appeal, or do you expect comments on here/​arguments by Said/​etc to affect the final outcome you decide on?

• bohaska 15 Aug 2025 3:50 UTC

  7 points

  0

  on: Bo­haska’s Shortform

  If we have already found features/​vectors for hallucinations, then why haven’t major AI companies tried shifting it downwards when deploying their AIs? Does reducing their strength decrease hallucinations? Is there a reason why using it in practice would not be helpful?

About 30% of Hu­man­ity’s Last Exam chem­istry/​biol­ogy an­swers are likely wrong

• bohaska 31 May 2025 9:30 UTC

  3 points

  3

  in reply to: gwern’s comment on: gw­ern’s Shortform

  This sounds like a whiteboard to me

• bohaska 22 May 2025 3:27 UTC

  7 points

  1

  on: Bo­haska’s Shortform

  Epoch AI’s new evaluation for Gemini 2.5 Flash Preview is broken.

  On their AI Benchmarking dashboard, the newest Gemini 2.5 Flash model is listed as having an accuracy of 4% ± 0.71% on GPQA Diamond, when Google’s official announcement lists it at over 80%, and when GPQA is a multiple-choice test with 4 options:

  

  It’s because of formatting issues. Helpfully, Epoch provides the logs from the evaluation, and the model just simply hasn’t been responding in the correct format.

  For example, if you look at the first sample from the logs, the correct answer is listed as “B”, but the model answered in Latex, $\boxed{B}$, so it was scored incorrect. There are plenty of other examples like this.

• bohaska 16 May 2025 8:08 UTC

  7 points

  0

  on: What Does It Mean to “Write Like You Talk”?

  “Write like you talk” depends on which language you are talking about.

  Take Arabic. Written Arabic and spoken Arabic has diverged enormously compared to written English and spoken English. Modern Standard Arabic (MSA) is the formal written language for books, newspapers, speeches etc. But no sane person speaks it. There are a lot of spoken dialects (like Egyptian, Levantine, Gulf Arabic, etc.). A speaker of different dialects may not understand other dialects or MSA, because all the vocabulary and grammar is different, which isn’t usually the case in English.
  
  Written and spoken English are similar to each other compared to most other languages.

• bohaska 13 May 2025 11:01 UTC

  1 point

  0

  on: AI Can’t Write Good Fiction

  I’ve tried using a different method for r1 to generate flash fiction: one sentence at a time. If a human writer wouldn’t write out a flash fiction story in one message, then AI shouldn’t, either. Here’s a result:

  She clocked in at 6AM, categorizing discards by residue: toothpaste-crusted wedding bands in Tier 4, melatonin vials from red-eyes in Tier 7, a child’s sock curled around her sister’s garnet earring (missing since the November her calls went unanswered).

  Room 312’s newlyweds left forensic poetry—dental floss strung bedpost to minibar, aspirin dust tracing slammed door trajectories. She logged these under Domestic Erosion, Subcategory: Honeymoon Phase.

  Room 214’s grid collapsed at the gel insoles—mint-green, bunioned, size 6 like her sister’s. The prescription (sertraline, 50mg) was dated three days after their last fight. She filed it under Unfinished Conversations, though the label peeled halfway.

  The businessman in 603 prayed, she’d assumed. But his trash betrayed her grids: glucose tabs bisecting train tickets, bloodied test strips where kneelers should’ve dented carpet. Her scrubbing split her cuticles, crimson streaking the sink’s rust.

  At dawn, she assembled her relics—unopened bills, an expired birth control foil (2019’s voicemail: static, then dial tone), lint rollers furred with 603’s hair. Each strand vibrated middle C, the note her sister had looped on the piano the night she vanished.

  Aspirin dust still gritted her palms. She pressed them to the window as dawn blued the glass—that bleached hue he’d called “motel dusk” while wrestling their tent zipper, his breath hot and futile against her neck.

  I think that this still has some imperfections, but I find that this method at least gives you an entirely different set of problems compared to the cliché output you describe.

• bohaska 1 May 2025 6:23 UTC

  12 points

  7

  in reply to: jimrandomh’s comment on: Early Chi­nese Lan­guage Me­dia Cover­age of the AI 2027 Re­port: A Qual­i­ta­tive Analysis

  I guess it’s just that the censors have not seen it yet.
  
  There’s a lot of situations where a smaller website doesn’t get banned e.g. Substack is banned in China, but if you host your Substack blog on a custom URL, people in China can still read it.

• bohaska 9 Apr 2025 14:12 UTC

  2 points

  1

  in reply to: ank’s comment on: Top Down & Physics-based Wor­ld­wide AI Alignment

  • Like Wolfram present a diffusion model as a world of concepts. But remove the noise, make the generated concepts like pictures in an art galley (so make 2D pictures stand upright like pictures in this 3D simulated art gallery), this way gamers and YouTubers will see how dreadful those model really are inside. There is a new monster every month on YT, they get millions of views. We want the public to know that AI companies make real-life Frankenstein monsters with some very crazy stuff inside of their electronic “brains” (inside of AI models). It can help to spread the outrage, if people also see their personal photos are inside of those models. If they used the whole output of humanity to train their models, those models should benefit the whole humanity, not cost $200/​month like paid ChatGPT. People should be able to see what’s in the model, right now a chatbot is like a librarian that spits quotes at you but doesn’t let you enter the library (the AI model).

  Okay, so you propose a mechanistic interpretability program where you create a virtual gallery of AI concepts extracted from Stable Diffusion, represented as images. I am slightly skeptical that this would move the needle on AI safety significantly, we already have databases like LAION which are open-source databases of scraped images used to train AI models, and I don’t see that much outrage over it. I mean, there is some outrage, but not a significantly large amount to be a cornerstone of an AI safety plan.

  gamers and YouTubers will see how dreadful those model really are inside. There is a new monster every month on YT, they get millions of views. We want the public to know that AI companies make real-life Frankenstein monsters with some very crazy stuff inside of their electronic “brains” (inside of AI models).

  What exactly do you envision that is being hidden inside these Stable Diffusion concepts? What “crazy stuff” is in it? I’m currently not aware of anything about their inner representations that is especially concerning.

  It can help to spread the outrage, if people also see their personal photos are inside of those models.

  It is probably a lot more efficient to show that by modifying the LAION database and slapping some sort of image search on it, so people can see that their pictures were used to train the model.

• bohaska 9 Apr 2025 13:05 UTC

  2 points

  0

  on: Top Down & Physics-based Wor­ld­wide AI Alignment

  Well, this assumes that we have control of most of the world’s GPU’s, and that we have “Math-Proven Safe GPUs” which can block the execution of bad AI models and only output safe AIs (how this is achieved is not really explained in the text), and if we grant this, then AI safety already gets a lot easier.
  
  This is a solution, but a solution similar to “nuke all the datacenters” and I don’t see how this outlines any steps that gets us closer to achieving it.

• bohaska 2 Apr 2025 2:38 UTC

  5 points

  0

  on: Bo­haska’s Shortform

  A helpful page to see and subscribe to all 31 Substack writers (out of 122 total) who were invited to LessOnline: https://​​lessonline2025invitedlist.substack.com/​​recommendations
  

  

• bohaska 2 Apr 2025 1:18 UTC

  10 points

  7

  in reply to: Aella’s comment on: Con­sider showering

  I guess this is another case of ‘Universal’ Human Experiences That Not Everyone Has