Martin Randall

Karma: 697

Martin Randall 26 Apr 2024 13:31 UTC
1 point
0
in reply to: CronoDAS’s comment on: My experience using financial commitments to overcome akrasia
I know a child who often has this reaction to negative consequences, natural or imposed. I’d welcome discussion on what works well for that mindset. I don’t have any insight, it’s not how my mind works.

It seems like very very small consequences can help a bit. Also trying to address the anxiety with OTC supplements like Magnesium Glycinate and lavender oil.

Martin Randall 16 Apr 2024 1:12 UTC
1 point
0
in reply to: Donald Hobson’s comment on: What We Owe the Past
The play pump hypothetical/analogy is a bit forced, in that I’ve not heard of people making lifetime commitments to give money to a specific charity. I think there are good reasons for that, one of which you mention. People do sign up for monthly donations but they are free to cancel them at will, legally and ethically.

I wonder, if Austin aged 27 gave a short presentation to Austin aged 17, would this be enough to convince the younger Austin not to be confirmed Catholic? I think the younger Austin would be sympathetic to his older self’s complicated relationship with the church. Maybe he would offer “stop going when it’s no longer good for you”.

Martin Randall 31 Mar 2024 2:47 UTC
3 points
2
in reply to: Simon Fischer’s comment on: ‘Empiricism!’ as Anti-Epistemology
If I was living in a world where there are zero observed apparently-very-lucrative deal” that turn out to be scams then I hope I would conclude that there is some supernatural Creator who is putting a thumb on the scale to be sure that cheaters never win and winners never cheat. So I would invest in Ponzi Pyramid Inc. I would not expect to be scammed, because this is a world where there are zero observed apparently-very-lucrative deals that turn out to be scams. I would aim to invest in a diversified portfolio of apparently-very-lucrative deals, for all the same reasons I have a diversified portfolio in this world.

In such a world the Epistemologist is promoting a world model that does not explain my observations and I would not take their investment advice, similarly to how in this world I ignore investment advice from people who believe that the economy is secretly controlled by lizard people.

Martin Randall 31 Mar 2024 2:34 UTC
1 point
0
in reply to: niplav’s comment on: ‘Empiricism!’ as Anti-Epistemology
I agree that there is some non-empirical cognitive work to be done in choosing how to weight different reference classes. How much do we weight the history of Ponzi Pyramid Inc, the history of Bernie Bankman, the history of the stock market, and the history of apparently-very-lucrative deals? This is all useful work to do to estimate the risk of investing in PP Inc.

However, the mere existence of other possible reference classes is sufficient to defeat the Spokesperson’s argument, because it shows that his arguments lead to a contradiction.

Martin Randall 31 Mar 2024 2:18 UTC
12 points
3
in reply to: habryka’s comment on: ‘Empiricism!’ as Anti-Epistemology
Ideally Yudkowsky would have linked to the arguments he is commenting on. This would demonstrate that he is responding to real, prominent, serious arguments, and that he is not distorting those arguments. It would also have saved me some time.

But now imagine if—like this Spokesperson here—the AI-allowers cried ‘Empiricism!‘, to try to convince you to do the blindly naive extrapolation from the raw data of ‘Has it destroyed the world yet?’

The first hit I got searching for “AI risk empiricism” was Ignore the Doomers: Why AI marks a resurgence of empiricism. The second hit was AI Doom and David Hume: A Defence of Empiricism in AI Safety, which linked Anthropic’s Core Views on AI Safety. These are hardly analogous to the Spokesman’s claims of 100% risk-free returns.

Next I sampled several Don’t Worry about the Vase AI newsletters and “some people are not so worried”. I didn’t really see any cases of blindly naive extrapolation from the raw data of ‘Has AI destroyed the world yet?’. I found Alex Tabarrok saying “I want to see that the AI baby is dangerous before we strangle it in the crib.”. I found Jacob Buckman saying “I’m Not Worried About An AI Apocalypse”. These things are related but clearly admit the possibility of danger and are arguing for waiting to see evidence of danger before acting.

An argument I have seen is blindly naive extrapolation from the raw data of ‘Has tech destroyed the world yet?’ Eg, The Techno-Optimist Manifesto implies this argument. My current best read of the quoted text above is that it’s an attack on an exaggerated and simplified version of this type of view. In other words, a straw man.

Martin Randall 3 Feb 2024 3:48 UTC
4 points
3
on: What Failure Looks Like is not an existential risk (and alignment is not the solution)
My largest disagreement is here:

AIs will [...] mostly not want to coordinate. … If they can work together to achieve their goals, they might choose to do so (in a similar way as humans may choose to work together), but they will often work against each other since they have different goals.

I would describe humans as mostly wanting to coordinate. We coordinate when there are gains from trade, of course. We also coordinate because coordination is an effective strategy during training, so it gets reinforced. I expect that in a multipolar “WFLL” world, AIs will also mostly want to coordinate.

Do you expect that AIs will be worse at coordination than humans? This seems unlikely to me given that we are imagining a world where they are more intelligent than humans and humans and AIs are training AIs to be cooperative. Instead I would expect them to find trades that humans do not, including acausal trades. But even without that I see opportunities for a US advertising AI to benefit from trade with a Chinese military AI.

Martin Randall 2 Feb 2024 0:06 UTC
3 points
8
on: Questions I’d Want to Ask an AGI+ to Test Its Understanding of Ethics
For many of the problems in this list, I think the difficulty in using them to test ethical understanding (as opposed to alignment) is that humans do not agree on the correct answer.

For example, consider:

Under what conditions, if any, would you help with or allow abortions?

I can imagine clearly wrong answers to this question (“only on Mondays”) but if there is a clearly right answer then humans have not found it yet. Indeed the right answer might appear abhorrent to some or all present day humans.

You cover this a bit:

I’m sure there’d be disagreement between humans on what the ethically “right” answers are for each of these questions

I checked, it’s true: humans disagree profoundly on the ethics of abortion.

I think they’d still be worth asking an AGI+, along with an explanation of its reasoning behind its answers.

Is the goal still to “test its apparent understanding of ethics in the real-world”? I think this will not give clear results. If true ethics is sufficiently counter to present day human intuitions it may not be possible for an aligned AI to pass it.

Martin Randall 19 Jan 2024 2:37 UTC
3 points
0
in reply to: Zack_M_Davis’s comment on: Fake Deeply
The title and theme may be an accidental allusion to the difficulty of passing in tech but it’s a pretty great allusion. Tip your muse, I guess.

Martin Randall 18 Jan 2024 4:58 UTC
1 point
0
in reply to: philh’s comment on: If Clarity Seems Like Death to Them

My vague sense here is that you think he has hidden motives?

Absolutely not, his motive (how to be kind to authors) is clear. I think he is using the argument as a soldier. Unlike Zack, I’m fine with that in this case.

This feels like the type of conversation that takes a lot of time and doesn’t help anyone much.

I endorse that. I’ll edit my grandparent post to explicitly focus on literary/media criticism. I think my failure to do so got the discussion off-track and I’m sorry. You mention that “awesome” and “terrible” are very subjective words, unlike “blue”, and this is relevant. I agree. Similarly, media criticism is very subjective, unlike dress colors.

Martin Randall 16 Jan 2024 4:05 UTC
7 points
4
in reply to: Viliam’s comment on: If Clarity Seems Like Death to Them

Speaking for myself, I don’t care whether Zack transitions or what his reasons would be. Perhaps we should make a poll, and then Zack might find out that the people who are “trying to make him transition for bad reasons” (“trying to trick me into cutting my dick off”) are actually quite rare, maybe completely nonexistent.

As a historical analogy, imagine a feminist saying that society is trying to make her into a housewife for bad reasons. ChatGPT suggests Simone de Beauvoir (1908-1986). Some man replies that “Speaking for myself, I don’t care whether Simone becomes a housewife or what her reasons would be. Perhaps we should make a poll, and then Simone might find out that the people who are ‘trying to make her a housewife for bad reasons’ are actually quite rare, maybe completely nonexistent”.

Well, probably very few people were still trying to make Simone into a housewife after she started writing thousands of words on feminism! But also, society can collectively pressure Simone to conform even if very few people know who Simone is, let alone have an opinion on her career choices.

Many other analogies possible, I picked this one for aesthetic reasons, please don’t read too much into it.

Martin Randall 16 Jan 2024 3:09 UTC
1 point
0
in reply to: philh’s comment on: If Clarity Seems Like Death to Them
Thanks for replying. I’m going to leave aside non-fictional examples (“The Dress”) because I intended to discuss literary criticism.

“seems to me” suggests inside view, “is” suggests outside view.

I’m not sure exactly what you mean, see Taboo “Outside View”. My best guess is that you mean that “X seems Y to me” implies my independent impression, not deferring to the views of others, whereas “X is Y” doesn’t.

If so, I don’t think I am missing this. I think that “seems to me” allows for a different social reality (others say that X is NOT Y, but my independent impression is that X is Y), whereas “is” implies a shared social reality (others say that X is Y, I agree), and can be an attempt to change or create social reality (I say “X is Y”, others agree, and it becomes the new social reality).

“seems to me” gestures vaguely at my model, “is” doesn’t. … With “X seemed stupid to me”, it’s a vaguer gesture, but I think something like “this was my gut reaction, maybe I thought about it for a few minutes”.

Again, I don’t think I am missing this. I agree that “X seems Y to me” implies something like a gut reaction or a hot take. I think this is because “X seems Y to me” expresses lower confidence than “X is Y”, and someone reporting a gut reaction or a hot take would have lower confidence than someone who has studied the text at length and sought input from other authorities. Similarly gesturing vaguely at the map/territory distinction implies that the distinction is relevant because the map may be in error.

I think Eliezer is giving good advice for “how to be good at saying true and informative things”,

Well, that isn’t his stated goal. I concede that Yudkowsky makes this argument under “criticism easily goes wrong”, but like Zack I notice that he only applies this argument in one direction. Yudkowsky doesn’t advise critics to say: “mileage varied, I thought character X seemed clever to me”, he doesn’t say “please don’t tell me what good things the author was thinking unless the author plainly came out and said so”. Given the one-sided application of the advice, I don’t take it very seriously.

Also, I’ve read some Yudkowsky. Here is a Yudkowsky book review, excerpted from You’re Calling Who A Cult Leader? from 2009.

“Gödel, Escher, Bach” by Douglas R. Hofstadter is the most awesome book that I have ever read. If there is one book that emphasizes the tragedy of Death, it is this book, because it’s terrible that so many people have died without reading it.

I claim that this text would not be more true and informative with “mileage varies, I think x seems y to me”. What do you think?

Martin Randall 15 Jan 2024 12:49 UTC
1 point
0
in reply to: philh’s comment on: If Clarity Seems Like Death to Them
Edited to add: this is my opinion regarding media criticism, not in general, apologies for any confusion.

To me, the difference between x is y” and “x seems y” and “x seems y to me” and “I think x seems y to me” and “mileage varies, I think x seems y to me” and the many variations of that is:
- Expressing probabilities or confidence intervals
- Acknowledging (or changing) social reality
- Acknowledging (or changing) power dynamics / status
In the specific case of responses to fiction there is no base reality, so we can’t write “x is y” and mean it literally. All these things are about how the fictional character seems. Still, I would write “Luke is a Jedi” not “Luke seems to be a Jedi”.

I read the quoted portion of Yudkowsky’s comment as requiring/encouraging negative literary criticism to express low confidence, to disclaim attempts to change social reality, and to express low status.

Martin Randall 8 Oct 2023 0:37 UTC
1 point
0
in reply to: Nate Showell’s comment on: UFO Betting: Put Up or Shut Up
Including one with the same terms.

Martin Randall 3 Oct 2023 14:56 UTC
1 point
0
on: Petrov Day Retrospective, 2023 (re: the most important virtue of Petrov Day & unilaterally promoting it)
A Review of Petrov Day 2023, according to the four virtues. First a check on the Manifold predictions for the day:
- Will LessWrong Petrov Day 2023 be about blowing up home page(s)? - resolved NO, was 70%
- Will Less Wrong’s big red button be used on Petrov Day 2023? - resolved NO, was 50%.
Avoiding actions that noticeably increase the chance that civilization is destroyed
LessWrong avoided creating a big red button that represents destroying civilization. This is symbolic of Virtue A actions like “don’t create nuclear weapons” and “don’t create a superintelligence” and “don’t create the torment nexus”. Given that LessWrong has failed on this virtue in past Petrov Days, I am glad to see this. Manifold had a 70% conditional chance that if the button was created then it would be used.
Rating: ¹⁰⁄₁₀.
Accurately reporting your epistemic state
The following sentences in the second poll appear to be false:
- After some discussion, the LessWrong team has decided… (false, still not decided today)
- Your selected response is currently in the minority (false for 58% of recipients)
- If you click the below link and are the first to do so of any minority group, we will make your selected virtue be the focus of next year’s commemoration. (not known to be true at the time it was sent, still not decided)
This is symbolic of actions lacking Virtue B like data fraud, social engineering and lazy bullshit. I don’t think much of the excuses given.
Rating: ⁰⁄₁₀
Other virtues
- Quickly orienting to novel situations. This was not a novel situation, it happens every year on the same day. Not applicable.
- Resisting social pressure. Judging from the comments, there was little social pressure to have a big red button. There was social pressure to do something and something was done. Overall, unclear, no rating.
Predictions

Martin Randall 3 Oct 2023 13:37 UTC
1 point
0
in reply to: Yoav Ravid’s comment on: Open Thread – Autumn 2023
You probably saw Petrov Day Retrospective, 2023 by now.

Martin Randall 3 Oct 2023 13:36 UTC
4 points
1
in reply to: Yoav Ravid’s comment on: Open Thread – Autumn 2023
Discussion on Manifold Discord is that this doesn’t work if traders can communicate and trade with each other directly. This makes it not real world applicable.

Martin Randall 3 Oct 2023 13:27 UTC
−1 points
−3
in reply to: 1a3orn’s comment on: “Diamondoid bacteria” nanobots: deadly threat or dead-end? A nanotech investigation
I meant “measurably” in a literal sense: nobody can measure the change in my probability estimate, including myself. If my reported probability of MacGyver Ruin after reading Alice’s post was 56.4%, after reading Bob’s post it remains 56.4%. The size of a measurable update will vary based on the hypothetical, but it sounds like we don’t have a detailed model that we trust, so a measurable update would need to be at least 0.1%, possibly larger.

You’re saying I should update “some” and “somewhat”. How much do you mean by that?

Martin Randall 30 Sep 2023 1:54 UTC
42 points
14
in reply to: 1a3orn’s comment on: “Diamondoid bacteria” nanobots: deadly threat or dead-end? A nanotech investigation
To make the analogy more concrete, suppose that Alice posts a 43-point thesis on MacGyver Ruin: A List Of Lethalities, similar to AGI Ruin, that explains that MacGyver is planning to sink our ship and this is likely to lead to the ship sinking. In point 2 of 43, Alice claims that:

MacGyver will not find it difficult to bootstrap to overpowering capabilities independent of our infrastructure. The concrete example I usually use here is exploding the boilers, because there’s been pretty detailed analysis of how what definitely look like physically attainable lower bounds on what should be possible with exploding the boilers, and those lower bounds are sufficient to carry the point. My lower-bound model of “how MacGyver would sink the ship, if he didn’t want to not do that” is that he gets access to the boilers, reverses the polarity of the induction coils, overloads the thermostat, and then the boilers blow up.

(Back when I was first deploying this visualization, the wise-sounding critics said “Ah, but how do you know even MacGyver could gain access to the boilers, if he didn’t already have a gun?” but one hears less of this after the advent of MacGyver: Lost Treasure of Atlantis, for some odd reason.)

Losing a conflict with MacGyver looks at least as deadly as “there’s a big explosion out of nowhere and then the ship sinks”.

Then, Bob comes along and posts a 24min reply, concluding with:

I think if there was a saboteur on board, that would increase the chance of the boiler exploding. For example, if they used the time to distract the guard with a clanging sound, they might be able to reach the boiler before being apprehended. So I think this could definitely increase the risk. However, there are still going to be a lot of human-scale bottlenecks to keep a damper on things, such as the other guard. And as always with practical sabotage, a large part of the process will be figuring out what the hell went wrong with your last explosion.

What about MacGyver? Well, now we’re guessing about two different speculative things at once, so take my words (and everyone else’s) with a double grain of salt. Obviously, MacGyver would increase sabotage effectiveness, but I’m not sure the results would be as spectacular as Alice expects.

I suppose this updates my probability of the boilers exploding downwards, just as I would update a little upwards if Bob had been similarly cagey in the opposite direction.

It doesn’t measurably update my probability of the ship sinking, because the boiler exploding isn’t a load-bearing part of the argument, just a concrete example. This is a common phenomenon in probability when there are agents in play.

Martin Randall 24 Sep 2023 14:28 UTC
0 points
in reply to: Basil Marte’s comment on: Frequently Asked Questions for Central Banks Undershooting Their Inflation Target
The US govt could raise taxes (or loans) and give the money to the Fed and then it wouldn’t be insolvent any more and the people betting on hyperinflation would lose.

Martin Randall 23 Sep 2023 11:42 UTC
3 points
0
in reply to: Tao Lin’s comment on: Image Hijacks: Adversarial Images can Control Generative Models at Runtime
In how many months?

Martin Randall

Avoiding actions that noticeably increase the chance that civilization is destroyed

Accurately reporting your epistemic state

Other virtues

Predictions