ScienceMon🔸

• ScienceMon🔸 14 Apr 2026 0:43 UTC

  2 points

  0 ∶ 0

  on: How scary is Claude Mythos? 303 pages in 21 minutes

  Your section “Mythos can hide its thoughts” reminds me a lot of the juncture in AI2027, where the fictional AI company must decide whether to press ahead with a model whose alignment properties were hazy. Something that’s unclear from your essay:

  Does Anthropic intend to correct the faulty safety training of Mythos (and Claude4.6)?

  This seems like a case where disclosure is not enough—they need to rectify the mistake in order to make these AIs as safe as possible. Especially since Claude4.6/​Mythos are likely to be used for the training & alignment of Claude6 and beyond.

• ScienceMon🔸 3 Mar 2026 0:05 UTC

  1 point

  0 ∶ 0

  in reply to: Jamie_Harris’s comment on: Slaugh­ter­bots are su­per un­pop­u­lar, and the US Govern­ment knows it

  The retaliation against Anthropic by labeling them a supply chain risk is indeed illegal. Congress authorized DoD to label companies as supply chain risks under very specific circumstances—the goal is to prevent foreign adversaries from infiltrating their components into our weapons systems. Hegseth flagrantly abused this power in a contract dispute with an American company. Anthropic has sued, and they will win in court.

  In the meantime, the question is whether American companies who do business with both DoD and Anthropic will comply with the illegal decree. Will those companies cut off their business with Anthropic?

  Having Congress stand up for the law will strengthen the spines of those companies. That is why I believe contacting your representatives is helpful on the margin.

Slaugh­ter­bots are su­per un­pop­u­lar, and the US Govern­ment knows it

• ScienceMon🔸 28 Feb 2026 19:44 UTC

  2 points

  0 ∶ 0

  in reply to: Aithir’s comment on: Pete Hegseth de­clares An­thropic a sup­ply-chain risk

  I appreciate the different perspectives @Aithir , and it takes guts to go against the crowd (I upvoted). What do you think EAs should do differently? Both inside and outside of Anthropic

• ScienceMon🔸 28 Feb 2026 3:31 UTC

  6 points

  2 ∶ 0

  on: Pete Hegseth de­clares An­thropic a sup­ply-chain risk

  I counted 13 lies. Not bad for 140 characters!

• ScienceMon🔸 9 Nov 2025 23:22 UTC

  6 points

  2 ∶ 0

  on: How and Why to Make EA Cool

  Welcome to the forum! For videos specifically, John & Hank Green have an enormous number of subscribers, and they’re pretty EA as far as I’m concerned. Just last week their vlogbrothers videos discussed treating tuberculosis in the Philippines.

  One thing I’d love to have is an EA Workout Playlist on Spotify. Music is cool. Fitness is cool. I’d love to have that in my daily routine.

ScienceMon’s Quick takes

• ScienceMon🔸 28 Sep 2025 14:27 UTC

  5 points

  2 ∶ 0

  on: ScienceMon’s Quick takes

  Iron supplementation is often important for vegans & vegetarians, especially if you also donate blood. Your doctor should measure your blood ferritin, not just hemoglobin. Both are important.

  Iron gluconate pills work fine for my wife, but they make me nauseous. Iron glycinate pills work better for me.

• ScienceMon🔸 25 Sep 2025 21:55 UTC

  13 points

  10 ∶ 0

  in reply to: SpeakClearly ’s comment on: Why you should eat meat—even if you hate fac­tory farming

  Welcome to the forum, @SpeakClearly. You make some good points here, but could you please engage in a friendlier way?

• ScienceMon🔸 25 Aug 2025 0:40 UTC

  24 points

  4 ∶ 0

  in reply to: Hugh P’s comment on: Hugh P’s Quick takes

  I’ve heard people say that the idea of an individual’s “carbon footprint” has actually harmed the cause of climate activists because it takes the emphasis onto personal behavior and off of policy change. Considering how amazingly successful policy advocacy for animal welfare can be, I worry that “welfare footprint” could be a step in the wrong direction.

• ScienceMon🔸 24 Apr 2025 11:21 UTC

  3 points

  0 ∶ 0

  in reply to: Jackson Wagner’s comment on: How con­fi­dent are you that it’s prefer­able for Amer­ica to de­velop AGI be­fore China does?

  Holy smokes this is a good answer! Do you know if anyone is trying to spread thinking about AI x-risk in China, at least among the engineers and intellectuals? I’m unsure about the tractability, but this seems super important & neglected since China still has a decent shot of developing AGI first.

  Several of the Statement on AI Risk signatories live in China, including a Dean at Tsinghua University. What can be done to further integrate them into the global AI x-risk community?

Save PEPFAR picnic

• ScienceMon🔸 1 Apr 2025 11:00 UTC

  1 point

  1 ∶ 0

  on: Sav­ing a life

  This is fantastic, phoebe. Keep it up!

• ScienceMon🔸 1 Mar 2025 23:19 UTC

  1 point

  0 ∶ 0

  in reply to: Joseph’s comment on: How con­fi­dent are you that it’s prefer­able for Amer­ica to de­velop AGI be­fore China does?

  Thanks for this feedback, Joseph. The bullet points I wrote were for sure overly simplified. I was trying to put myself into the shoes of an open-minded Chinese citizen, who has no doubt absorbed more pro-China propaganda than either of us has.

  “How would you try to convince an open-minded Chinese citizen that it really would be better for America to develop AGI first?”

  If it is possible to convince Chinese AI engineers that “losing the race” is in their best interest, then that would be a huge win for everyone. It would give the West more breathing room to develop AGI safely.

[Question] How con­fi­dent are you that it’s prefer­able for Amer­ica to de­velop AGI be­fore China does?

• ScienceMon🔸 18 Feb 2025 2:26 UTC

  9 points

  6 ∶ 0

  on: The abrupt end of USAID-funded clini­cal tri­als is profoundly unethical

  There is a word that educated liberals need to learn or relearn: evil.

  Flagrant disregard for human life is evil.

• ScienceMon🔸 5 Feb 2025 3:13 UTC

  1 point

  0 ∶ 0

  in reply to: Tristan Cook’s comment on: Do short AI timelines de­mand short Giv­ing timelines?

  Thanks for those links, Tristan! It felt a bit like @Jackson Wagner’s comment was scolding me directly:

  The idea that a neartermist funder becomes convinced that world-transformative AGI is right around the corner, and then takes action by dumping all their money into fast-acting welfare enhancements, instead of trying to prepare for or influence the immense changes that will shortly occur, almost seems like parody.

  Why do you believe that capital won’t be useful?

[Question] Do short AI timelines de­mand short Giv­ing timelines?

• ScienceMon🔸 24 Oct 2024 20:28 UTC

  5 points

  0 ∶ 1

  on: Miles Brundage re­signed from OpenAI, and his AGI readi­ness team was disbanded

  OpenAI is nothing without its people.

• ScienceMon🔸 26 Aug 2024 12:59 UTC

  7 points

  1 ∶ 0

  on: Ca­reer FOMO. How to exit?

  It sounds like we’re around the same age, both a few years out of a PhD (mine was in bio). I’m happy to email/​talk 1-on-1 with you about this:

  1. Do not continue down the academic path. Your mind and body are clearly telling you to stop! Instead, start applying for jobs as you wind down your ongoing projects.

  2. Probably don’t start a business right now. Not unless you have a major technical edge in a lucrative area and a suite of business-relevant skills (I couldn’t infer this from your post).

  3. If your internships were at large/​established companies, I would be unsurprised that they went poorly. There are stark cultural differences between academia and the corporate world.

  4. Consider joining a startup. Culturally, this will be a smoother transition than to big corporate. You will be paid much more than you are as a postdoc. The people around you will generally be happier than your academic colleagues. And you’ll build skills that are relevant to starting your own business one day.

  Academia is wonderful in many ways, but it teaches people that life is linear, which is a damn lie. Life isn’t linear! You have decades ahead of you that will be filled with personal growth and bringing happiness to other people.