robryk @robryk@qoto.org

@sgf

Let me start by saying that I will have blind spots here, because this touches things that are close to one of the few really virtue-based/deontological rules that I have: I will not try to affect people's thinking/beliefs without being explicit about what I'm doing. (I have it partially for "good" reasons, partially because doing otherwise feels very uncomfortable, so it's easier for me to just not do that.)

I think that trying to be empathetic, understanding your interlocutor's position are good approaches in general (not because they allow you to convince people, but because they sometimes lead to you discovering something new[^]). I think that doing that _specifically because you want to convince them_ would not work if you told them that you're doing that in order to be more likely to convince them. If I were to try to imagine not obeying the no-underhandedness rule, I'd expect that the interlocutor would find my attention to them to be artificial, but this is very unconfident given I have little experience with that.

I think when trying to understand one's interlocutor's position is combined with willingness to change own mind if this provides convincing evidence, then this is a ~technique that is actually biased towards truth. If it's not combined with that, I think one cannot actually employ it convincingly and while not being underhanded. I'm not sure if that version is biased toward truth.

-

I think that we have been thinking of different things as "argument-based methods". I would've included Socratic questioning[^^] in those -- because it relies on (implicit if done well) creation of arguments, but I now expect you wouldn't (because there's no argument being provided to the person being convinced). This makes me think that we largely disagreed in phrasing and not substance, so I'll address what I think are disagreements in substance below, but they are smaller than I expected.

(The rough reason for me ~equating truth-biased with argument-based is that -- in areas other than maths where everything is usually argument based anyway -- ISTM that any truth-biased approach must include a way to notice that it's misapplied, and that seems to require exchanging reasoning and pointing out flaws in it.)

> If someone is out into conspiracy theories, I'd rather get rid of the conspiracy theory than start with improving critical thinking. To my mind, that can come later.

Intuitively I'd expect that doing the former is not very useful, because the mechanism that caused them to believe in this falsehood will probably cause them to believe in multiple others. The only(?) way I can see of achieving former in a durable fashion without latter is if the person gets convinced that their reasoning is faulty and that they should pick an authority and follow them, which I'm not sure is actually better (because it feeds noise into positive feedback loops of people being considered authorities).

> From another angle, I'm not sure which techniques you're thinking of that you think should help defend against conspiratorial thinking, but ironically I've found "rationalist" techniques to be very poor for this.

Examples:
- explicitly thinking about things that a particular belief rests on (so, if any of them turned out false, the belief should be reevaluated),
- "if the argument doesn't rely on property X, it should hold for situations that differ only in property X",
- the concept of bisecting arguments to find the implication you disagree with (and the concept of formulating arguments in a way that makes that easier)[^^^].

I'm not sure that I've addressed everything that you've said and I made some assumptions about what you think above, so please do point out where I overlooked something or erred.

[^] Sadly, when that happens too rarely, one is conditioned away from those approaches.

[^^] I'm not proposing it's reasonable in context; just providing an example for definitions

[^^^] I'm at a total loss when someone disbelieves in modus ponens, though.

@dejanfreiburg

To be fair, ww2 bombs are found quite often during construction work, so there's some positive correlation between the first and last.

@jacob @TechConnectify @lukechannings

Let's say a pilot in an airliner gets on the PA and says "the rudder is not on fire"[^]. Each and every passenger was pretty sure that was true before the pilot said it, so _if the message conveys only its literal meaning_ the passengers should basically ignore it. And yet, I would expect this to be a significant message for the passengers: the pilot choosing to make such an announcement is very surprising and thus brings lots of information.

IMO this example supports the statement that sadly, unless we are in some context where there are strong social norms about being explicit about each and every part of the message one wants to convey, one cannot assume others will read only the literal meaning of the sentences that are published. The only situations in which I've seen such strong norms is working together on a problem that requires precision and correctness.

[^] h/t to John Cleese et al, whose "How to Irritate People" is the source of the example

@m0bi13

Jestem zawiedziony punktem 6. Porównuje on płace minimalną w 2019 z tą w 2023, ignorując inflację (odpowiednik tej z 2019 w 2023 to byłoby ~3000PLN). Na podstawie tego porównania i braku wzrostu bezrobocia krytykuje on krytyke podniesienia płacy minimalnej do wartości, która w dzisiejszej walucie wynosiłaby ~5800PLN.

@jonmsterling

Do you mean explicit branches in code when you say branching? If so, I'm somewhat confused: I can get nondeterministic behaviour from a concurrent system with no explicit branches (example: thread A does write(false to x); write(true to x); thread B does return read(x);).

(I agree that interleaving of operations is not a correct way to describe an execution of a concurrent program and that (even though for many memory models one can use a linear order on something other than operations) a partial order where each thread is totally ordered and which satisfies some other memory-model-specific requirements is both a correct and natural representation of an execution.)

@sgf

It does make sense and is at least somewhat related. I'll probably say something more detailed in a day or a few, once I've thought about it some more and when it's not late at night :)

Why do we say that batteries/energy storage is something that stores and releases energy?

A device that is able to convert some amount of heat into zero-entropy energy is ~just as useful as a battery of the same capacity. It's obvious that such a device can exist (a thermos with some amount of heat capacity inside at a temperature lower than environment + a heat engine is such a device), and it has at least one significant advantage over a battery: it is not necessarily able to release a significant quantity of energy on failure/destruction.

One can easily create such batteries in that exact way (by using a thermos and heat engine). My rough upper bound is that the highest "energy" density one could get without phase transition would be ~2MJ/kg[*] (by using gaseous hydrogen as the medium), which is a bit better than current batteries (but is an upper bound that ignores any practicalities of handling hydrogen, inefficiencies other than thermodynamically necessary, and the weight of infrastructure). I've looked at a few phase transitions (e.g. ice/water or liquid/gaseous nitrogen) and they don't seem to be able to give anything even close to the value for hydrogen (or, for that matter, mere Li-Ion batteries).

Are there other ways to store negative entropy? I imagine that chemical ones "should" exist, but I have terrible intuition for entropy changes across chemical reactions to even know where to start looking.

[*] I've taken half of the energy that would be needed to heat hydrogen from ~0K to ~room temp (half because efficiency will change linearly between 1 and 0 as the temperature of the hydrogen increases).

tl;dr Are there any better ways to store negative entropy without storing energy than storing very cold hydrogen? Is there some sort of fundamental limitation in play?

#thermodynamics

@sgf

One thing I perhaps didn't make clear: I think that nearly always convincing a person of one particular fact is ~unimportant compared to improving their likelihood of acquiring convictions that are true in the future.

@sgf

I don't really have one; it's one of the things I can't effectively do unless the person cooperates quite a bit (by being at least somewhat curious).

> Bear in mind that the "non-argument-based methods" don't do away with facts, they just use them differently.

Hm~ I'm not sure which methods you mean exactly. For example, claiming that existence of an anecdote proves the general statement that the anecdote would be an example of kinda relies on facts, but nevertheless is extremely bad way to change one's beliefs to more truthful ones.

If you know of methods that don't "teach"/reinforce false implications, are more strongly "truth-aligned" than the anecdote examples above and yet are not argument based, I'd really want to know about them.

@TechConnectify

Thank you, it's very nice to know about this split. Do you know if the music side itself has the same split?

I also wonder about the split across authors of videos I see. ISTM, given that Premium is a fixed price subscription, that watching _more_ videos decreases the fraction that I send to a particular author. I'm not sure what incentives it creates (looking at first order effects, it would seem to me that YT is incentivized to decrease the amount of videos Premium subscribers watch; that's probably not true for PR reasons if no other, so my naive views of incentives here would probably be wrong), but intuitively it feels very weird.

@sgf

Hm~ I don't know of the exact conversation you've had, and you've summarized your opinion here very briefly, so what I say might be obvious/beside the point/unsurprising.

The non-argument-based methods usually give less (or even no) comparative advantage to true claims over false claims. The more trusted and popular they become in a given group of people, all else being equal that group would usually become worse at agreeing on true statements (and individuals in that group are likely to be worse at distinguishing truth from falsity individually).

Now, one might have different reactions to that. One of them is to never use non-argument-based methods, because if one uses them to argue for true things people trust conclusions provided via those methods more. Another similar one is to object to others doing that.

I don't think these approaches are obviously wrong (i.e. that the assumptions that would lead to them being sensible are self-contradictory or very obviously incompatible with how the world could look like with a different but stable culture). They might be (I'm not good at predicting what people do around forming beliefs) totally impractical, but that in itself is a nontrivial question.

@TechConnectify

Do you know what's the order of magnitude of the ratio of the revenue split?

@jonmsterling

Then I'm curious which departure from that you think makes that model an oversimplification: more lax memory models, fairness constraints on the scheduler, something else I'm not thinking of, or some combination?

@jonmsterling What do you mean by powerdomain semantics?

@quasirealsmiths Hm~ I don't get why. I often see Calvin's dad giving false explanations meant satirically in a way that (at least to me) would feel discouraging if I actually wanted to know the answer.

@eta IIRC I saw ones where every second row was localized.

@isomer it's not clear what you want to weigh with

@rotopenguin @mw1cgg @delroth

Doesn't the USB-C plug cut the signal lines first when you pull it out? If so, you don't need the plug to be able to withstand disconnecting power lines in normal operations: you could have at least one end cut off power lines internally when data (well, probably CC as opposed to the rest of data) lines go.