Posts

What should OpenAI do that it hasn't already done, to stop their vacancies from being advertised on the 80k Job Board? 2024-10-21T13:57:30.934Z
The 27 papers 2024-05-30T08:46:39.587Z
GPT-4o is out 2024-05-13T18:33:53.890Z
Everything Wrong with Roko's Claims about an Engineered Pandemic 2024-02-22T15:59:08.439Z

Comments

Comment by WitheringWeights (EZ97) on Understanding Shapley Values with Venn Diagrams · 2024-12-09T13:25:09.219Z · LW · GW

I was teaching myself bits of cooperative game theory and this is the clearest explanation I've found so far. I think it's a nice complement to this one.

Comment by WitheringWeights (EZ97) on "It's a 10% chance which I did 10 times, so it should be 100%" · 2024-11-20T21:24:45.449Z · LW · GW

I thought to myself... "That looks familiar..."

 

It is also very similar to the formula for calculating the compound interest rate. 
Just swap the minus with a plus and the function tends to e: after all, compounding interest rates was how the constant got known in the first place.

Comment by WitheringWeights (EZ97) on Information vs Assurance · 2024-10-23T10:12:34.320Z · LW · GW

Me: I dunno, probably around 9 pm. [At this point, I’ve merely offered some information; I think most people would not interpret this as an assurance, and would not blame me much if I show up to the party at 8:30 or 10:00 or even skip it altogether.]


Assuming the conversation doesn't delve further into this, if I were your friend I'd actually be very surprised if you didn't show up. The question 'At what time are you going?' assumes that you're going, however uncertain the details. If you wish to convey the idea of 'you might not see me at all' your answer should explicitly include 'but I might not go' because without that clause you're agreeing to attend, at least at some point.

To be clear, I agree with the gist of the piece. I just find it funny how even such a short convo could lead to a quite dramatic misunderstanding.

Comment by WitheringWeights (EZ97) on You don't know how bad most things are nor precisely how they're bad. · 2024-08-05T14:54:21.563Z · LW · GW

An anecdote from which I drew a similar conclusion to yours:

About ten years ago I went with my dad to a music hall where a local marching band was playing. I play a few instruments, have a solid grasp of music theory, etc..., but I'm no professional, while he has 'average Joe' music training.
I found the concert to be genuinely painful to listen to: entire sections not in tune with each other, very poor dynamics (brass way too loud, woodwinds barely audible), melodies all over the place, sudden tempo changes etc...

After the last piece, the audience asked for the encore. I figured it had to do with social pressure/convention/kindness. Before I could say anything, my dad proceeded to extol the band's musicianship and high level of the performance, assuming I too had enjoyed the experience. 

I was (and still am) baffled.



 

Comment by WitheringWeights (EZ97) on Sticker Shortcut Fallacy — The Real Worst Argument in the World · 2024-06-12T15:31:05.032Z · LW · GW

“Your argument is fallacious because X is not a central category member. And it’s not a central category member because…I say so?”

In my view, part of what makes the non-central argument a fallacy is the ad hoc use of the 'overly restrictive definition'.

Whoever argues that "MLK is a criminal" with the intent of instilling the negative connotation of the term is unlikely to apply the same standard everywhere. 

One in that case could reply that anyone who ever opposed any non-democratic regime and was found guilty of sedition/instigation/etc.... is/was also a criminal. 

If the proponent of the argument disagrees, we fall into a contradiction. If in favour of the extended use, the original concept - criminal - is stripped off its stereotypical baggage.

Comment by WitheringWeights (EZ97) on Soviet comedy film recommendations · 2024-06-10T19:48:39.453Z · LW · GW

I believe Gaidai's Ivan Vasilievich Changes His Profession (Иван Васильевич меняет профессию) also deserves an honorable mention, I am pretty sure it should be available with English subtitles.

Comment by WitheringWeights (EZ97) on Two easy things that maybe Just Work to improve AI discourse · 2024-06-09T12:00:02.867Z · LW · GW

Assuming that the goal is to 'raise the sanity waterline', I would recommend against engaging on most social media platforms, except for promoting content that resides outside the media circle (e.g. a book, blogpost, paper, etc...).
 

Comment by EZ97 on [deleted post] 2024-06-08T08:45:07.980Z
Comment by WitheringWeights (EZ97) on Help me to become "less wrong" · 2024-05-30T10:10:14.690Z · LW · GW

The comments below are just my impressions, hope you'll find them useful.
Generally: frame the problem, define the terms, cite the sources, ask specific questions.

For example:

Statement A: Two of three children survived the day. Is this good or bad?
Answer: Neither, as it is an "is" statement. The moral judgment depends on the context, such as a woman considering aborting triplets.

Statement B: Two of three children were murdered today. Is this good or bad?
Answer: This is bad because murder is considered morally wrong, making it an "ought" statement.


Statement B is a 'is' statement too. 'Is it good or bad' is by definition an ought statement.

For instance, some argue that an AI would be "dumb" to prioritize making paperclips over human life. However, this presupposes that the AI subscribes to the idea that human life ought to be prioritized, ignoring the possibility that the AI may not necessarily hold this view.

Who argues what? Some argue that the earth's flat. This doesn't make it worth discussing. 
Here it seems that you're borderline strawmanning a poorly written and argued imaginary position (similarly to the section on the 'gender debate').
Also, the objection you raise misses the mark of the "AI would be dumb" argument. If I say that an intelligent system would assign high value to human life (because of training data, design, etc....) this would not amount to an 'ought' statement.

Therefore, both examples fail to describe the subject of discussion.

The Submerged Premise Problem occurs when participants in a debate operate on fundamentally different foundational beliefs or assumptions that are not explicitly stated or acknowledged. These hidden assumptions shape their arguments, leading to misunderstandings and preventing meaningful progress in discussions.

I think a better framing of the picture you paint (especially about the 'gender debate') is given in this article.

The correct thesis would be that there are no genders

I don't understand how 'no gender' would be the 'correct thesis' for either side of the debate. Not that the position is nonsensical, but why would it derive from the premises?

The discussions around the mid-2010s made these examples more prevalent, ironically affecting society's ability to understand the alignment problem in AI.

How do you establish causation?

In contrast, the culture wars before this era, such as the debate around intelligent design, were fought on more solid ontological grounds. There was a yield in these debates, with many prominent conservatives eventually accepting the theory of evolution. However, in today's social media-driven landscape, discussions often sink into the mud, lacking the same level of ontological clarity and resolution.

How do you draw a comparison of ontological ground solidity across different public debates? These statements ought to be backed with some sources and miss a lot of important context in order to draw inferences.

Looking forward for your opinion.

I personally wouldn't know how to contribute since the terms are so open ended. I would ask specific questions (after clarifying all the above).

I think the post would be fine as a first draft but requires some additional work.

Comment by WitheringWeights (EZ97) on A Gentle Introduction to Risk Frameworks Beyond Forecasting · 2024-04-12T08:51:11.706Z · LW · GW

Strongly upvoted, neat overview of the topic.
I especially like the academic format (e.g. with the sources clearly cited), as well as its conciseness and breadth.

Comment by WitheringWeights (EZ97) on The Best Tacit Knowledge Videos on Every Subject · 2024-04-05T10:46:28.054Z · LW · GW

“American financial criminal and businessman. Shkreli is the co-founder of the hedge funds Elea Capital, MSMB Capital Management, and MSMB Healthcare, the co-founder and former CEO of pharmaceutical firms Retrophin and Turing Pharmaceuticals, and the former CEO of start-up software company Gödel Systems, which he founded in August 2016” (Wikipedia).


It should at least be mentioned that Shkreli is a convicted fraudster.

Comment by WitheringWeights (EZ97) on Everything Wrong with Roko's Claims about an Engineered Pandemic · 2024-02-24T11:15:22.961Z · LW · GW

Most of the sources he cites are also referred to in my article, which of his points do you find the strongest that are not already addressed here?

Comment by WitheringWeights (EZ97) on Everything Wrong with Roko's Claims about an Engineered Pandemic · 2024-02-24T11:14:06.245Z · LW · GW

I cite and contextualise the news article in my post (see second half of sec. 'Misrepresentation of evidence')
Ebright is cited in Butler's news post on Nature also referred to by Roko. The post is a brief summary of a debate on the usefulness of GoF research, and makes some other points, among which:

-Barich arguing that the project was funded because not so risky as to fall under the moratorium;

-"Although almost all coronaviruses isolated from bats have not been able to bind to the key human receptor, SHC014 is not the first that can do so. In 2013, researchers reported this ability for the first time in a different coronavirus isolated from the same bat population." 
Natural coronaviruses were already able to bind to human receptors, that's arguably why GoF research was conducted in the first place.

In light of this, I don't think the news article is worth a significant update toward the lab-leak hypothesis: if anything, it shows that it was known since 2013 that coronaviruses could infect humans without host animals.



 

Comment by WitheringWeights (EZ97) on Less Wrong automated systems are inadvertently Censoring me · 2024-02-21T20:26:10.002Z · LW · GW

You can still write posts, it doesn't look like brute-force manufactured consensus to me. Your original post got over 200 karma which seems pretty high for a censorship attempt (whether intentional or not).

Comment by WitheringWeights (EZ97) on My cover story in Jacobin on AI capitalism and the x-risk debates · 2024-02-13T23:21:43.611Z · LW · GW

An addendum to this: Marinetti embraced an early form of Italian fascism ('sansepolcrino') that already by 1925 had been in fact disavowed by fascist leaders and that was pretty much antithetical to fascism as most people intend it.
Here the 1919 Fascist Manifesto from Wikipedia, I am familiar with the original document and it's correct. Not to be confused with the 1925 Manifesto of the Fascist Intellectuals.

Politically, the Manifesto calls for:

  • Universal suffrage with a lowered voting age to 18 years, and voting and electoral office eligibility for all ages 25 and up;
  • Proportional representation on a regional basis;
  • Voting for women;
  • Representation at government level of newly created national councils by economic sector;
  • The abolition of the Italian Senate (at the time, the Senate, as the upper house of parliament, was by process elected by the wealthier citizens, but were in reality direct appointments by the king. It has been described as a sort of extended council of the crown);
  • The formation of a national council of experts for labor, for industry, for transportation, for the public health, for communications, etc. Selections to be made of professionals or of tradesmen with legislative powers, and elected directly to a general commission with ministerial powers.

In labor and social policy, the Manifesto calls for:

  • The quick enactment of a law of the state that sanctions an eight-hour workday for all workers;
  • A minimum wage;
  • The participation of workers' representatives in the functions of industry commissions;
  • To show the same confidence in the labor unions (that prove to be technically and morally worthy) as is given to industry executives or public servants;
  • Reorganization of the railways and the public transport sector;
  • Revision of the draft law on invalidity insurance;
  • Reduction of the retirement age from 65 to 55.

In military affairs, the Manifesto advocates:

  • Creation of a short-service national militia with specifically defensive responsibilities;
  • Armaments factories are to be nationalized;
  • A peaceful but competitive foreign policy.

In finance, the Manifesto advocates:

  • A strong extraordinary tax on capital of a progressive nature, which takes the form of true partial expropriation of all wealth;
  • The seizure of all the possessions of the religious congregations and the abolition of all the bishoprics, which constitute an enormous liability on the Nation and on the privileges of the poor;
  • Revision of all contracts for military provisions;
  • The revision of all military contracts and the seizure of 85 percent of the profits therein.
Comment by WitheringWeights (EZ97) on Lying to chess players for alignment · 2023-10-25T20:33:14.423Z · LW · GW

Hi!

I'm rated between 1500 and 1700 on lichess, I'd be happy to take part in the game in whatever role.

Comment by WitheringWeights (EZ97) on Enforcing Far-Future Contracts for Governments · 2023-10-05T09:54:11.420Z · LW · GW

Hey, thanks a lot for bringing up the topic, I think it deserves more attention in general and I don't get some of the objections raised in the comments.

I'm not sure I understood how investors would value the perpetuity/annuity.
What I mean is that the least they can expect is to receive the annuity for a minimum of 50 years (following the example), so they would probably go with a valuation of this sort so to already discount the case in which the state reneges on its pledge. When this happens (because of MPs or justices) there's no one left to actually care about the law being passed, and no financial consequences for the state since investors had already discounted the possibility of reneging.

Another point is targeted default not affecting the credibility of a country, given the nature of the debt, but I won't go into that since it has already been argued in the comments.

An issue I see is that policy makers are not that committed to reducing public debt, since it doesn't affect them personally and can always kick the can down the road/pass the buck/blame it on past legislatures. 

This very much depends on the nature of the reform and basic cost-benefits, but I believe that lack of incentive and the targeted default not being seen as catastrophic weaken the enforcement mechanism.

I think you should write a longer piece explaining more in detail your proposal in light of the comments, I'd personally be happy to read more on this.

Comment by WitheringWeights (EZ97) on Far-Future Commitments as a Policy Consensus Strategy · 2023-09-24T11:36:42.489Z · LW · GW

The main problem I see here lies in implementing an institutional mechanism that ties future governments to bills passed decades before.
What I mean is that future governments are likely to be able to further delay/ignore past 'bills for the future' because a legislature is only bound to act without breaching the respective constitution (usually interpreted by a constitutional court), its bylaws (decided internally by MPs) and the law (written by MPs).
This implies that any law that does not enjoy constitutional ranking can be just as easily repealed by any future legislature using the same procedure that was used to pass the bill to the future.

I would also say that over decades I would expect most countries to experience changes that would likely stem the enacting of a past law (for instance, major constitutional reforms/more or less violent institution changes that expunge legal orders and create new once).

Note that I keep the wording vague because this issue applies to most existing countries.