MichaelDickens

• MichaelDickens 20 Feb 2026 18:10 UTC

  5 points

  1 ∶ 0

  in reply to: NickLaing’s comment on: Effec­tive Altru­ism Has Very Good Epistemics

  I think using made-up-ish numbers can be better than not doing it at all—as long as there’s no impression of rigor. We don’t have to claim rigor!

  In my experience, here’s how these things go ~100% of the time:

  • Authors make up some numbers, and they include about a dozen caveats about the limitations of their model.

  • Readers ignore all the caveats and accuse them of claiming to be rigorous, even though they claimed no such thing.

  AI 2027 is a great example of this.

• MichaelDickens 18 Feb 2026 1:09 UTC

  8 points

  1 ∶ 0

  in reply to: MichaelDickens’s comment on: Is bench­mark­ing AI ca­pa­bil­ities pos­i­tive EV?

  A relevant question I’m not sure about: for people who talk to politicians about AI risk, how useful are benchmarks? I’m not involved in those conversations so I can’t really say. My guess is that politicians are more interested in obvious capabilities (e.g. Claude can write good code now) than they are in benchmark performance.

• MichaelDickens 18 Feb 2026 1:05 UTC

  9 points

  1 ∶ 0

  on: Is bench­mark­ing AI ca­pa­bil­ities pos­i­tive EV?

  Know when to sound alarm bells

  What is the situation where people coordinate to sound alarm bells over an AI benchmark? I basically don’t think people pay attention to benchmarks in the way that matters: a benchmark comes out demonstrating some new potential danger, AI safety people raise concerns about it, and the people with the actual power continue to ignore them.

  Thinking of some historical examples:

  • Anthropic’s findings on alignment faking should have alerted people that AI is too dangerous to keep building, but it didn’t.

  • Anthropic/​OpenAI recently finding that their latest models may cross ASL-4 /​ CBRN thresholds should’ve been an alarm bell that they can’t release those models, but they released them anyway and nobody could do anything to stop them.

• MichaelDickens 18 Feb 2026 0:57 UTC

  4 points

  0 ∶ 0

  on: Is bench­mark­ing AI ca­pa­bil­ities pos­i­tive EV?

  Increases the speed of ai development

  This is the biggest con by far. There are at least two mechanisms by which it could happen:

  • Make AI capabilities growth more apparent, incentivizing investment.

  • Help AI companies learn how to build more powerful AI systems.

• MichaelDickens 14 Feb 2026 2:18 UTC

  9 points

  1 ∶ 0

  on: Long-term risks from ide­olog­i­cal fanaticism

  We focus on intentional deaths because they are most revealing of terminal preferences, which in turn are most predictive of future harm.

  Why is an ideological desire to kill the outgroup more likely to influence the long-term future than (say) a desire to preserve wild-animal suffering for aesthetic reasons?

  Considering that wild animal suffering is currently somewhere around 10^3 to 10^15 times worse than any ideologically-driven tragedy has ever been, by default I’m far more concerned about the former.

• MichaelDickens 8 Feb 2026 1:00 UTC

  13 points

  5 ∶ 0

  in reply to: Ben_West🔸’s comment on: Ben_West’s Shortform

  Ah yes, this supports my pre-conceived belief that (1) we cannot reliably ascertain whether a model has catastrophically dangerous capabilities, and therefore (2) we need to stop developing increasingly powerful models until we get a handle on things.

• MichaelDickens 30 Jan 2026 2:26 UTC

  12 points

  2 ∶ 3

  in reply to: Vasco Grilo🔸’s comment on: Un­falsifi­able sto­ries of doom

  A donor wanted to spend their money this way; it would not be fair to the donor for Eliezer to turn around and give the money to someone else. There is a particular theory of change according to which this is the best marginal use of ~$1 million: it gives Eliezer a strong defense against accusations like

  If they suddenly said that the risk of human extinction from AGI or superintelligence is extremely low, in all likelihood that money would dry up and Yudkowsky and Soares would be out of a job.

  I kinda don’t think this was the best use of a million dollars, but I can see the argument for how it might be.

• MichaelDickens 29 Jan 2026 16:17 UTC

  15 points

  3 ∶ 1

  in reply to: Dave Banerjee 🔸’s comment on: Un­falsifi­able sto­ries of doom

  Copying from my other comment:

  The reason Eliezer gets paid so much is because a donor specifically requested it. The express purpose of the donation was to make Eliezer rich enough that he could afford to say “actually AI risk isn’t a big deal” and shut down MIRI without putting himself in a difficult financial situation.

  (I don’t know about Nate’s salary but $235K looks pretty reasonable to me? That’s less than a mid-level software engineer makes.)

  Edit Feb 2: Apparently the donation I was thinking of is separate from Eliezer’s salary, see his comment

• MichaelDickens 29 Jan 2026 16:14 UTC

  12 points

  3 ∶ 2

  in reply to: Yarrow Bouchard 🔸’s comment on: Un­falsifi­able sto­ries of doom

  MIRI pays Eliezer Yudkowsky $600,000 a year.

  I believe this is because a donor specifically requested it. The express purpose of the donation was to make Eliezer rich enough that he could afford to say “actually AI risk isn’t a big deal” and shut down MIRI without putting himself in a difficult financial situation.

  Edit Feb 2: Apparently the donation I was thinking of is separate from Eliezer’s salary, see his comment

• MichaelDickens 27 Jan 2026 3:37 UTC

  4 points

  0 ∶ 0

  in reply to: Vasco Grilo🔸’s comment on: If AI al­ign­ment is only as hard as build­ing the steam en­g­ine, then we likely still die

  I spent a few days thinking about this but I struggled to come up with a bet structure that I was confident was good for both of us. The financial implications of this sort of bet are complicated. I didn’t want to spend more time on it so I’ll punt on this for now but I will keep it in the back of my mind in case I come up with anything.

• MichaelDickens 20 Jan 2026 19:14 UTC

  2 points

  0 ∶ 0

  Error

  The value
    NIL
  is not of type
    SIMPLE-STRING
  when binding #:USER-ID162

• MichaelDickens 19 Jan 2026 22:04 UTC

  6 points

  0 ∶ 0

  in reply to: SummaryBot’s comment on: Re­turn Stacked Funds: A New Way to Get Leverage

  This summary is accurate.

• MichaelDickens 18 Jan 2026 21:28 UTC

  2 points

  0 ∶ 0

  in reply to: PeterMcCluskey’s comment on: Re­turn Stacked Funds: A New Way to Get Leverage

  Do you have an opinion about what maturity is best to short?

  (My first thought is that if you have a view about what interest rates will do over the next 5–10 years, then you should short 5–10 year bonds. But I’m not sure that’s right.)

Re­turn Stack­ing Funds: A New Way to Get Leverage

• MichaelDickens 14 Jan 2026 20:59 UTC

  4 points

  0 ∶ 0

  in reply to: Vasco Grilo🔸’s comment on: If AI al­ign­ment is only as hard as build­ing the steam en­g­ine, then we likely still die

  Are you assuming the deployment of ASI will be analogous to an omnipotent civilisation with values completely disconnected from humans suddenly showing up on Earth?

  Something like that, yeah.

  However, that would be very much at odds with historical gradual technological development shaped by human values.

  ASI would have a level of autonomy and goal-directedness that’s unlike any previous technology. The case for caring about AI risk doesn’t work if you take too much of an outside view, you have to reason about what properties ASI would have.

• MichaelDickens 14 Jan 2026 13:49 UTC

  5 points

  0 ∶ 0

  in reply to: yujing’s comment on: yu­jing’s Quick takes

  Donating at the end of the year.

  I used to donate mid-year for the reasons you gave. The last couple years I donated at the end of the year because the EA Forum was running a donation election in early December, and I wanted to publish my “where I’m donating” post shortly before the donation election, and I don’t want to donate until after I’ve published the post. But perhaps syncing with the donation election is less important and I should publish and donate mid-year instead?

If AI al­ign­ment is only as hard as build­ing the steam en­g­ine, then we likely still die

• MichaelDickens 8 Jan 2026 4:31 UTC

  4 points

  0 ∶ 0

  on: Amer­i­can effec­tive al­tru­ists should con­sider poli­ti­cal donations

  A concern with donating to political candidates is that every candidate who puts major focus on AI safety is a Democrat; funding their campaigns risks dialing up polarization of AI safety. I’m pretty uncertain about how big a deal this is.

• MichaelDickens 8 Jan 2026 4:28 UTC

  21 points

  8 ∶ 0

  on: Public in­tel­lec­tu­als need to say what they ac­tu­ally believe

  Eliezer comes to mind as a positive example:

  • He’s pretty explicit about explaining why he believes what he believes.

  • He gives the true reason for his beliefs, not the reason he expects to be most persuasive, even if the true reason is “weird”.

• MichaelDickens 25 Dec 2025 19:54 UTC

  22 points

  8 ∶ 0

  on: “In­fluence from within” on AI doesn’t work, so bye bye

  I’ve been a fool trying to influence people who are on the AI industry’s money and glory payroll. I’m going to take my own advice now, write you off, and focus on the moral majority who wants to protect the world.

  I’ve donated $30,000 to PauseAI. Some of your past posts played a role in that, such as The Case for AI Safety Advocacy to the Public and Pausing AI is the only safe approach to digital sentience. I don’t think writing off people like me is a good idea.

  You should all be ashamed of your complicity in bringing about potentially world-ending technology.

  I am literally donating to PauseAI. I don’t think you are being fair. I fully agree that some EAs are directly increasing x-risk by working on AI development, and they should stop doing that. I don’t think it’s fair to paint all of us with that brush.