All of Michael Wiebe's Comments + Replies

Should you "trust literatures, not papers"?
I replicated the literature on meritocratic promotion in China, and found that the evidence is not robust.

Do vaccinated children have higher income as adults?
I replicate a paper on the 1963 measles vaccine, and find that it is unable to answer the question.

New replication: I find that the results in Moretti (AER 2021) are caused by coding errors. The paper studies agglomeration effects for innovation (do bigger cities cause technological progress?), but the results supporting a causal interpretation don't hold up.

What was the effect of reservists joining the protests? This says: "Some 10,000 military reservists were so upset, they pledged to stop showing up for duty." Does that mean they were actively 'on strike' from their duties? It looks like they're now doing grassroots support (distributing aid).

4Yoav Ravid4mo
They only said that they would not come to reserve duty training, but have repeatedly said in case of a real emergency they will all show up immediately.

They did indeed stop going to reserve. However, basically 99% of them who were of relevant age enlisted now. It's not actually in our cultural DNA not to show up in emergency. Brothers In Arms also called for everyone to enlist immediately (Hebrew, sorry).

Yeah, I do reanalysis of observational studies rather than rerunning experiments.

But isn't it problematic to start the analysis at "superhuman AGI exists"? Then we need to make assumptions about how that AGI came into being. What are those assumptions, and how robust are they?

I strongly agree with this objection. You might be interested in Comprehensive AI Services, a different story of how AI develops that doesn't involve a single superintelligent machine, as well as "Prosaic Alignment" and "The case for aligning narrowly superhuman systems". Right now, I'm working on language model alignment because it seems like a subfield with immediate problems and solutions that could be relevant if we see extreme growth in AI over the next 5-10 years. 

Why start the analysis at superhuman AGI? Why not solve the problem of aligning AI for the entire trajectory from current AI to superhuman AGI?

1Lone Pine2y
EY's concern is that superhuman AI might behave wildly different than any lesser system, especially if there is a step where the AI replaces itself with something different and more powerful (recursive self-improvement). In machine learning, this is called "out-of-distribution behavior" because being superintelligent is not in the 'distribution' of the training data -- it's something that's never happened before, so we can't plan for it. That said, there are people working on alignment with current AI, such as Redwood Research Institute, Anthropic AI and others.

Also came here to say that 'latter' and 'former' are mixed up.

In particular, we should be interested in how long it will take for AGIs to proceed from human-level intelligence to superintelligence, which we’ll call the takeoff period.

Why is this the right framing? Why not focus on the duration between 50% human-level and superintelligence? (Or p% human-level for general p.)

So it seems very likely to me that eventually we will be able to create AIs that can generalise well enough to produce human-level performance on a wide range of tasks, including abstract low-data tasks like running a company.

Notice how unobjectionable this claim is: it's consistent with AGI being developed in a million years.

If you're loss averse, the expected value could easily be negative: cost(voting for wrong candidate) > benefit((voting for right candidate).

I was astonished to find myself having ascended to the pantheon of those who have made major contributions to human knowledge

Is this your own evaluation of your work?

Yes. The basic situation is that I figured out how the methods that Charles Spearman used to discover IQ can be used to shed a lot of light on many different psychology and sociology questions. This is what I was implicitly getting at in my sequence of posts on my Speed Dating Project, though I did a poor job contextualizing the results. IQ is by far the most robust construct to come out of psychology research, so this could in principle revolutionize social science (with a huge amount of work by many talented people). Some people would say that psychology researchers used the methodology to discover the Big 5 Personality traits among other things, but the constructs that they developed are relatively weak because they haven't utilized the full power of Spearman's framework. See also Sarah Constantin's post A Yardstick for Smell: Thinking in PCA (she preempted me in recognizing the potential of the methods 4.5 years before I did, but I didn't fully understand what she was getting at at the time). It's difficult to make such claims without coming across as grandiose and/or arrogant, so I should emphasize that I think that many LWers would be capable of doing such research with the right background and style of research, if they took Paul Graham and Steve Jobs' advice. The great mathematician Alexander Grothendieck said: [...]

If the "tear apart the stars" prophecy just refers to Harry harvesting the stars for resources, then Voldemort looks really stupid for misinterpreting it.

Without going back to The Good Book, it was perfectly clear to Harry at the end "tear apart the stars" prophecy could just refer to Harry harvesting the stars for resources. ok, I'll peak. Does that mean that Quirrell knew the exact prophecy? And Dumbledore knew it too. Were Quirrell and Dumbledore basing his worries about Harry destroying the stars on the prophecy? Shouldn't they just have asked? "Hey, what could it mean if someone is supposed to "tear apart the very stars in heaven"? But Dumbledore then tells it to him in his letter anyway. And Quirrell discusses his apprehension that Harry will destroy the sun, but doesn't mention the prophecy at the time. It seemed like there was a lot of heartburn over something Harry could have cleared up in a minute.
Star-lifting isn't well known even among sci-fi fans in 2015. In 1991 you would have to be a fairly serious space-geek/sci-fi fan to know about it.

Now Hermione learns Patronus 2.0 and destroys Azkaban. So both the Boy-Who-Lived and the Girl-Who-Revived can kill dementors. Sounds like "surviving/defeating Voldemort" is a plausible cover for explaining the origin of the ability to destroy dementors.

With Dementors out of the way, the cost of telling people the secret of the True Patronus becomes a lot lower.

Isn't Harry saying this to Draco after Draco has been obliviated? Draco has no idea what Harry's talking about.

Shouldn't Harry have fallen to his knees twenty seconds earlier, if he originally heard/saw the explosion via Voldie-simulcast?

What do you mean? He narrated the whole sequence before the explosion, and fell to his knees at the moment Voldemort supposedly died, which is coincident with the explosion. I don't see a problem, let alone one that would be fixed by shifting the narrative back 20 seconds.
He was claiming to be dipping into Voldemort's working memory, not experiencing it directly - that's why he knew what happened to Dumbledore. Anyway, that could plausibly provide a level of insulation.
We do not know at what speed does information travel through the Riddle link.
Now that may be just what Draco needs to realize that something's wrong with Harry's story. Maybe Harry didn't know the timing well enough to fall at the correct second and had to listen for the crack.

"Harry, let me verify that your Time-Turner hasn't been used," said Professor McGonagall.

"LOOK OVER THERE!" Harry screamed, already sprinting for the door.

Harry can just claim to have already used it that day for an innocuous purpose, like studying or something. Sure, McGonagall could accuse him of stupidity because that leaves him unprepared for an emergency, but pleading guilty to stupidity is easy. (Well, easier, anyway.)
A story for the masses is necessary and this doesn't appear to be a bad stab at one. Harry can always bring trusted others on board by telling them what actually happened. He might have actually done that already and this is their plan. How much time did Harry have to do stuff before needing to show up anyhow (40m? 50m?)? Also, Prof. McGonagall is terrible at faking anything so telling her the truth before this seems like a bad idea.
Also, won't anyone go back with a time-turner to check Harry's version of events? Also, won't Snape know?
When does TT reset to unused? At midnight?

I think there is a very good chance that McGonagall is worried enough about Hermione, about Dumbledore, ... and won't check Harry's Time-Turner. She's not much into multilevel plots, she's smart, but too honest to live in constant suspicion.

Like she didn't think much about the consequences at the "war level" when she offered the fealty oath to Hermione back then in the Wizengamot.

She sees Harry in pain and traumatized, and she sees the immediate chaos, it wouldn't be very "in character" for her to suddenly suspect Harry and probe him li... (read more)

0Ben Pace9y
I suppose she's the courageous hero.

Important: QQ's earlier parseltongue-spoken plans for Harry to become ruler of the world were said before he heard the 'tear apart the stars' prophecy. So it appears V changed his mind after hearing the prophecy.

That's entirely reasonable. Parseltongue translates your thoughts to your voice, it doesn't bind your actions.

If knowledge of the True Patronus can prevent people from being able to cast Patronus 1.0, is there a way for knowledge of the True Patronus to harm Voldemort?

If what he said back in front of the Dementor is true, Voldie can't cast the Patronus regardless.

Does Quirrell have the Resurrection Stone? If so, that's 3/3 Deathly Hallows (invisibility cloak and elder wand).

He had it when he learnt what it was, and he would hardly have let himself lose it afterwards.

When Harry first entered the room wearing his cloak, he looked into the mirror and saw only the reflection. Now he is again looking into the mirror while cloaked.

Dumbledore is in the mirror. Quirrell, from 104:

I saw the Headmaster missing... but for all my magic can tell me... he could be in another... realm of existence

Yup. Once you think of it, he couldn't leave that one out. There has to at least be a big nod, though the actual murder won't necessarily happen.

But unless you bought Draco Malfoy's latest theory that Professor Sprout had been assigning and grading less homework around the time of Hermione being framed for attempted murder, thereby proving that Professor Sprout had been spending her time setting it up, the truth remained unfound.

So Imperiused-Sprout is Hat and Cloak?

Ironic that The Watson actually got the right answer...
Somehow I can't see Sprout having Hat-and-Cloak's eloquence, and I doubt you can Imperius someone to be significantly more eloquent than their usual self.

Yes. But two minutes before that he was thinking of taking Cedric, and then we get a cut scene to him sneaking about in the hallway with Lesath. That implies that Cedric might still be in play, otherwise we probably would've gotten a short sentence or two on why he chose Lesath over Cedric.

Cedric, who may or may not have a time turner, could quite possibly show up.

We are nearing the end of the school year, after all.

Edit for clarity: referring to the curse on the Defense Against the Dark Arts teaching position.

In the future, poverty reduction EAs might also focus on economic, political, or research-infrastructure changes that might achieve poverty reduction, global health, and educational improvements more indirectly, as when Chinese economic reforms lifted hundreds of millions out of poverty.

I'd like to see more discussion of economic growth and effective altruism. Something that can lift hundreds of millions of people out of poverty is something that should definitely be investigated. (See also Lant Pritchett's distinction between linear and transformative philanthropy.)

Regarding Harry's lack of surprise, isn't it odd that he puts no effort into wearing the expression of someone who has no idea about Hermione's body being missing?

Remember that Harry had just been hastily awoken long before his accustomed time. It's not unreasonable for Harry to be behaving a little bit awkwardly, and it certainly isn't enough of a tell for Dumbledore to draw any conclusions. What does seem to be a bit of a tell is his strange behavior around the ring; he seems to deliberately create tension before the ring is verified in order to, apparently, play for sympathy afterwards.
Sarajevo server nymphs retreat. Shaven tart Marjory perseveres. Majesty's pervert overran rheas. Jeremy's transverse vapor earth. Trojan's preservers tame Harvey. Jester's revery proves amaranth. Jerry's pervert overate shamans. Rajahs' poverty nerve streamers. Majors preserve earthy servant. ... and a lot of other things.

Is there a page that lists all of the unresolved hints/clues in MoR? For example, Remembrall-like-a-sun, Bacon's diary, etc.

This list has been compiled on reddit some time ago. It goes up to chapter 85. Some things have been resolved since then.

He had vanished from where he was standing over the Weasley twins and come into existence beside Harry; George Weasley had discontinously teleported from where he was sitting to be kneeling next to his brother's side

What's going on here? Is it just that Harry isn't paying attention to what's happening around him?

Prediction: Harry will attempt to learn Obliviation, use his Time-Turner to go back to before, and attempt to mess with his own head to save Hermione while preserving his own experience of events. This is more likely to not work than work.

No, the abruptly-ended and grammatically-incorrect sentence preceding this passage indicates actual discontinuity:

"Dumbledore wasn't being very cooperative, and in any case this was several minutes after the critical location within Time"

Notice the lack of punctuation. The end of this sentence has been lopped off, and deliberately. Eliezer Yudkowsky does not make careless punctuation errors.

Yeah, it sounded like a first person perspective of Harry-in-shock to me.

This is interesting. From the end of Ch. 89:

Unseen by anyone, the Defense Professor's lips curved up in a thin smile. Despite its little ups and downs, on the whole this had been a surprisingly good day

From Ch. 46, after Harry destroys the dementor:

I must admit, Mr. Potter, that although it has had its ups and downs, on the whole, this has been a surprisingly good day.

Every day that Harry kills something is a good day, of course.

Do you think that the standards given in the OP are too demanding? Not demanding enough?

If you have read 10 papers on a political question, all the papers are concurring, and they represent the entire body of literature on that question, then 90% can be warranted. I strongly suspect that if you do a fair review of any question there will be a tremendous amount of disagreement, however, and a well-calibrated observer would rarely approach 90% without strong pre-existing ideological biases subverting their estimation processes. I have read quite a few macroeconomics papers, and there are very few non-trival things that I would assign a 90% probability to, and almost by definition those aren't the sort of things people talk about in "political discussions." The more you read the more humble you become with respect to our level of knowledge in the social sciences, especially if you are literate in the hard sciences. If you expect your knowledge to cash out in terms of predictions about the world, we don't know much. This is why if you ask an economics professor almost any question of substance, they will provide you with a lengthy docket of caveats.

Good point. Do you think the ideological Turing-capability requirement helps to mitigate this danger, and if so, how much does it help?

Academic papers are what get's published, not what's true. The difference is particularly pronounced for political topics.

Right, it's a necessary condition, not a sufficient one.

It's not a necessary condition. Academic papers are regularly mistaken either due to methodological limitations, bad statistical methodology, publication bias, stretching of conclusions and a host of other factors. The fact that information has been published is evidence that it contains true information, but is not even necessarily strong evidence. Keep in mind that in the field of cancer research, which is on much firmer quantitative footing than political science "research," researchers were unable to replicate 89% of studies.

As a tool for combating privileged questions, what about consciously prioritizing which issues you spend time thinking about?

I think of it the other way around: combating privileged questions is a tool for consciously prioritizing which issues you spend time thinking about.

Both strategies might end up producing the same outcome. Define a "Certified Political Belief" as a belief which satisfies the above standards. In my own case, I don't actually have any strong political beliefs (>90% confidence) which I would claim are Certified (except maybe "liberal democracy is a good thing").

In fact, a good exercise would be to take your strongest political beliefs, actually write down which academic articles you've read that support your position, and then go do a quick test (with a third-party referee) to see whether you're ideologically Turing-capable. This sounds like a good way to get feedback to help you calibrate.

I interpreted the signals as "this woman is interesting," yet when I got to know those woman, I was not actually >interested in their personality. I put a lot of effort into fixing this miscalibration, and I think it was worth the effort.


I was interested in intellectual women, and somehow got it in my mind a that certain kind of contrarian signalling in women was evidence that they were intellectual.

Thus, I tried to spend more time with that type of woman, both as friends and potential partners. But that type of woman didn't find me very interesting. Several times, particular individuals were painfully careless with my emotions.

Eventually, I realized that (1) those types of people didn't find me interesting, and (2) I didn't actually find those types of people interesting. So whenever ... (read more)

What's a good term for "being able to pass an ideological Turing test"? (Being able to pass an ITT is related to being able to argue both sides of a debate, being able to accurately explain your opponent's position, being able to summarize the strongest counterargument to your position, etc.)

Following the original analogy, is there a term for "a machine that's able to pass a Turing test"? My googling didn't turn up anything. But if there was ("a machine is called Turing-(blank) if it can pass a Turing test"), then it seems we could adapt it fairly easily to the ITT: someone is ideologically Turing-(blank) if they can pass an ITT.

Any suggestions to fill in the blank?

0Michael Wiebe11y

On a related topic, Pinker has a very useful discussion of the case for and against open discussion of dangerous (non-technological) ideas. (Mindkiller warning)

A person is said to exhibit rational irrationality when it is instrumentally rational for him to be epistemically irrational. An instrumentally rational person chooses the best strategies to achieve his goals. An epistemically irrational person ignores and evades evidence against his beliefs, holds his beliefs without evidence or with only weak evidence, has contradictions in his thinking, employs logical fallacies in belief formation, and exhibits characteristic epistemic vices such as closed-mindedness. Epistemically irrational political beliefs can rei

... (read more)
I think this quote might have the analysis backwards. Politicians are not irrational for spouting irrational nonsense - because that is what voters want to hear. I'm not sure if that is accurately described as "epistemically irrational" because some of the politicians probably know what the correct answers are. None of that creates incentives on voters to be epistemically irrational - except for game-theoric reasons. There certainly are costs to voters being epistemically irrational (assuming one believes there are meaningful differences between the political parties - which may not be the local consensus.
I wish we would reconsider the upvote/downvote mechanics on LW.

I would recommend skipping the section on political correctness. I do think the first two sections give a good lesson on how a little reason can be a dangerous thing.

Looks like he got hoist by his own petard.

This article seems relevant: "Clever sillies: Why high IQ people tend to be deficient in common sense."

The author argues that high IQ people solve problems by using abstract reasoning instead of evolved common sense. Moreover, general intelligence is mainly useful for solving evolutionarily novel problems, and can actually be a hindrance for problems which were a regular part of the evolutionary environment (for example, social situations). Hence, when facing problems where humans have evolved behavioral responses, smart people who apply abstract reasoning and override common sense often end up doing silly things.

Unfortunately, it's by Bruce Charlton. I've noticed that whenever this hypothesis comes up, it seems to be solely used as a political cudgel to attack liberals - which means I trust the paper as far as I can throw it.

(Why is the 'clever silly' idea always used to attack things like diversity, and not equally abstract and historically unprecedented shibboleths of the right like untrammeled free markets?)

Load More