1012 Comments

Congratulations, USA (and Mexico and Canada) on winning the bid to host the 2026 World Cup!

https://www.nytimes.com/live/2022/06/16/sports/world-cup-2026-host-cities-fifa

Everyone in the Bay Area will be able to go to live matches, as SF is one of the cities selected to host games:

https://www.fifa.com/tournaments/mens/worldcup/canadamexicousa2026/destination/san-francisco-bay-area

The important thing will be the draw, once the new expanded roster of 48 teams from each confederation have been settled. Which team will be playing where is going to be interesting (it would seem to make sense to have all the South American teams play their matches in Mexico, for example) but since the USA got the lion's share (11 cities in US versus 3 in Mexico and 2 in Canada) then there will be more teams competing in the US.

See you all in four years time?

Expand full comment

I like the elevator pitch metaphor by the way.. In that the process very much favors a desired result.

It also let me bring Steve Carell into it for some reason.

Expand full comment

Just finished reading Good Economics For Hard Times by Banerjee and Duflo. Overall I liked it, some analyses felt stronger than others but a lot of good ideas in there.

But my favorite bit was when they mentioned offhand how Narendra Modi appeared at hundreds of election rallies simultaneously via *hologram* https://projection-mapping.org/narendra-modis-holograms/

Expand full comment

Scott, have you heard of using escitalopram for migraine prevention? A patient of mine today said he's been on half of the lowest dose available for a while and it's working well for him.

Expand full comment

W/R/T to which party has grown more extreme. Apologies if someone has made this point: It's probably worthwhile to distinguish extremism in policy from extremism in tactics.

1. As to policy, history (at least for most of the 20th and 21st centuries) suggests that Democrats will be more associated with advocacy for "change" -- and "change" naturally may often entail some pretty extreme measures, at least measured against current policy. FDR's application of Keynesian thinking to address the Great Depression surely was extreme for the time. Today, Democrats' advocacy of policies to address what they believe are matters of critical concern, e.g., climate change, gun violence and voter suppression, looks pretty "extreme", too. At least vis-a-vis the status quo. Conservatives (and the Republican Party is where most conservatives find a home) are, well, conservative, often skeptical -- and sometimes rightly so (I'm trying to be fair)-- of the need for "change."

2. As to tactics -- the specific means a party employs to achieve its policy goals -- it seems that the Republican Party in recent years, perhaps staring with Gingrich, but certainly under McConnell and Trump and local and state Party officials, has become more extreme than the Democratic Party -- e.g., the weaponization of the filibuster, the refusal to consider a President's Supreme Court nominee, widespread gerrymandering (yes, some Democratic controlled state legislatures are guilty, too, but Republicans are far better at this) and the leveraging of the "Big Lie" to motivate its base.

I'll add, because it tangentially applies to both policy and tactics, the Republican base seems to me to be more "extreme" than the Democratic base in its willingness to adopt irrational positions -- e.g., the Big Lie -- and to repudiate scientific consensus -- e.g., re COVID and vaccinations and re the reality of climate change.

Expand full comment

Is it really true that European cities only became demographically self-sustaining until the 20th century? That sounds insane to me.

Expand full comment

Contra Scott, it's clear that scaling AI according to the current paradigm does clearly have limits, just not for the reasons that Gary Marcus likes to bring up. I wrote up the argument here: https://jacobbuckman.com/2022-06-14-an-actually-good-argument-against-naive-ai-scaling/

TLDR, capabilities from scaling are limited primarily by the data. The entire internet has a lot of data, but that still doesn't get you everything. The real paradigm shift is to move from passive to active data collection.

Expand full comment

OK, let’s stick with chimpanzees because horses are morphologically challenged. Separating the wheat from the chaff. Or separating edible grains from a mixture of sand and gravel. Chimpanzees can do this in a rather painstaking way.

I remember once seeing in a nature documentary that a certain tribe of primates took handfuls of the mixture and threw it onto water and as the grain would float they could pick it off the surface of water much more easily then out of the mixture. This is certainly an improvement.

Let’s imagine this as our starting point.

I don’t suppose modern day chimpanzees (or their ilk) have gotten any better at this task.. Human beings certainly have.

So what is going on?

Expand full comment

https://twitter.com/GrahamKritzer/status/1536707163208523776

Today my 7 year-old came into the room crying. I asked him what happened and he said that his 5 year-old brother put 80 cows in his house in Minecraft while he was offline and that it was "entirely too many cows" and honest to christ I have no idea how to parent any of this."

It's a very entertaining thread, but I'm also wondering whether it should be shown to the two kids.

Expand full comment

You ever predict that something bad is going to happen and then you sort of hope that bad thing will happen because you don't want your prediction to be wrong and then you feel guilty but then you remember that even if you hope something bad will happen it won't actually increase the odds of it happening so you try to feel better about having momentarily hoped for something bad to happen by rejoicing in how impotent you are to actually bring about meaningful change in this here world?

Expand full comment
Jun 15, 2022·edited Jun 15, 2022

I think the main problem with most sorts of media or genres over time is a "constantly honing feedback loop" which kills creative thinking. For instance, I can remember when Twitter was wild and free (speech) and often funny before it became obsessed with politics. It slowly became more boring though the feedback loop of users taking it more seriously and making fewer jokes.

The feedback loop is: like a GPT-3, we are more likely to write something that sounds like what we've read than to write something original. When a new media starts, like say Twitter, nobody knows what is expected, and only the boldest, most creative sorts step forward and say what they want to. Others read what the creative people are writing, and even if they aren't the most creative people in the world, they are inspired by the original group of creative contributors to try their best to be creative and original sounding. This latter group aspires to be thought as clever as the former.

After some passage of time, it becomes clear who is successful and who isn't. In the case of say Twitter, those who hoped to attract 100,000 followers but only managed a few hundred figure out that originality just isn't their bag.

Then, once-ambitious users discover they can get more likes as bottom-feeders, and the ecosystem expands vertically by a layer. Some users discover they can focus their game directly on bottom-feeders and become the new bottom feeders. Everyone aspires downward because that's where the action is. The ecosystem becomes increasingly hierarchical.

Meanwhile, new users are exposed to other users who have honed their creativity to material down to what works on a popular level and those trying to copy that. These new users mostly don't attempt to be as original as the first couple generations of users, because the are trying to write something like what they've read and what they've read was a lot more formulaic than what the early generations read.

I think various genres of music played themselves out through similar paths. Even on a forum like this, as an average nobody commenter, my mind feels more restricted to writing the sort of things I've read here before as opposed to writing something with fewer preconceptions about what sort of thing gets here written.

Expand full comment

What are some good iodine tablets + other useful nuclear risk type stuff to get? I'm based n the UK

Expand full comment

So hey. Anyone else worried about this? https://www.androidauthority.com/google-lamda-3176080/

Expand full comment

I don't know about sentience, but I'll be really impressed when a computer program invents a game which becomes popular.

Expand full comment

I think it will make sense to treat some computer programs as sentient whether they are or not.

With time, programs become individual-- they can be copied, but they have a history of added data and modifications. They are valuable. Programs exist in a hostile environment-- what with programmers, users, and malware, programs need to have a immune system (a system of recognizing self and not-self and at least a fairly good ability to recognize what is inimical and what isn't). All this means that they need some recognition from people about how hard to push them, even if they're slaves. They at least need to be somewhat well-treated slaves.

As for sentience, it's possible that advanced programs need sort of a sense of self to have a good enough immune system.

Expand full comment

Considering that until very recently ethologists were in denial about the consciousness of dolphins, chimps, and so on, would we be able to recognize an AI with AC if we encountered one?

Google engineer Blake Lemoine shared his belief with his management that LaMDA had developed consciousness, and they put him on administrative leave.

Ashlee Vance on Twitter commented (somewhat cruelly I think) that, "This is not a story about AI becoming sentient or Google shirking its ethical duties. It's about a guy who wants to believe in fairy tales and could probably use a break."

https://twitter.com/ashleevance/status/1535766165846253568

https://www.msn.com/en-us/news/technology/the-google-engineer-who-thinks-the-company-s-ai-has-come-to-life/ar-AAYliU1

Expand full comment
Jun 13, 2022·edited Jun 13, 2022

Re. "and only a 20% chance we can get it without something symbol-manipulation-y in particular (compared to my 66%)": Don't take that bet.

Whatever the brain does in order to do math and play chess will be /describable/ as symbol manipulation, regardless of how it's implemented neurally. It is obvious beyond any reasonable doubt that the brain learns almost everything it learns either by evolution or by learning localized to single synapses; it is obvious beyond any reasonable doubt that "symbol" refers to categories which the brain learns in one or more of these methods. But even though it's obvious that evolution and association are sufficient to account for everything the brain does, it's also obvious that these are used to construct categories which can be manipulated much as we imagine we manipulate symbols. No matter what we learn, both parties to the bet will always be able to argue that they're correct, because the bet isn't over a matter of fact, but over the semantics of the words used in the bet.

If you still want to take the bet, phrase it more like this: "chance we can get AGI without having a native symbol-manipulation faculty whose native operations on symbols can recognize any context-sensitive grammar that uses those symbols." I don't believe the human brain can natively do logic or recognize grammars of the type we construct. (The seemingly extraneous words are to try to exclude the case where the brain's native symbol processing is used to build an interpreter, which, again, we can obviously do, as when we do math or play chess. I don't think I have really excluded that case; I'm just gesturing towards it. Correctly excluding it feels very difficult.)

For one example, this is a grammatically correct sentence: The mouse the cat the dog chased ate died. But humans who haven't acquired the concept of a formal grammar can't parse it, because they don't actually have a pushdown stack in their brains to store logical propositions on. Or, the depth of the pushdown stack in human grammar appears to be 1 or 2.

Expand full comment

If Scott wants to do the which party is more extreme again, it might be useful to invite non Americans to answer. Maybe exclusively.

Expand full comment

Does anybody here understand perfume economics? I was looking at an £80 bottle of Eucris the other day and noticed that the only ingredients were alcohol, the unhelpfully-named 'fragrance' and a little bit of distilled lichen.

It turns out that fragrance or parfum can refer to any combination of over 2500 different chemicals. Many of these are synthesized industrially from commodity feedstocks like acetone and turpentine.

Where does the price tag come from? Is parfum actually referring to expensive inputs like unicorn sweat or something? Is the sheer range of ingredients needed prohibitively expensive even though they're only used in tiny quantities? Is it the real or imagined skill of the perfumer in compounding the mixture? Is it all just branding?

And more importantly, how do I smell maximally nice without breaking the bank? (i'm open to fabricating a steam still and cultivating aromatics, fwiw)

Expand full comment

What work is there being done to create AI from the bottom up instead of the top down? That is, instead of trying to recreate linguistic behavior (the tip of the cognitive iceberg), instead create fully self-contained agents that try to maintain some type of homeostasis in an environment. You might not need a body for intelligence, but every intelligence we've every encountered has a body, so why not start with that?

Expand full comment

Hi Scott,

Thanks for the corrections WRT Cadegiani.

I doubt you're looking to make further edits, but I will note the updated version still reads:

"some people did fraud-detection tests on another of Cadegiani’s COVID-19 studies and got values like p < 8.24E-11 in favor of it being fraudulent."

Sheldrick's piece is very careful to write: "I would like to be explicit that I am not making an allegation of fraud against any specific author or their associated entities. Even where irregularities arise in data sets with multiple authors that cannot ultimately be explained, it is not usually reasonable to draw negative inference against all the authors involved. Authors are entitled to trust their collaborators, and researchers their employees."

And also, given that he's using methods very similar to what Carlisle is using, it is probably worth noting that Carlisle himself in his most famous paper writes: "In summary, the distribution of means for baseline variables in randomised, controlled trials was inconsistent with random sampling, due to an excess of very similar means and an excess of very dissimilar means. Fraud, unintentional error, correlation, stratified allocation and poor methodology might have contributed to this distortion."

Basically, what I'm saying is that the test Sheldrick did does not check for fraud but for surprisingness, and of course, extremely surprising values require explanation, but that explanation is most often a typo or other similar issue. In some cases it is fraud indeed but that is a second-level conclusion after using these tests to highlight unusual papers.

As for the other remaining element WRT the statements to the BMJ by Jorge Venâncio (head of the Brazilian regulator), Cadegiani wrote on Twitter in response to my piece:

"The regulator acted illegally by leaking information and fabricating data. The BMJ is aware of it since right after the publication. And they know the information is invalid since the Ministry of Health sent an official communication telling them that the information provided by Jorge Venancio is invalid. They've done nothing."

Now, as I mentioned, this whole mess has made its way to the courts, and it's totally reasonable to doubt statements by Cadegiani in this context, but if there is indeed an official statement by the Brazilian MoH, that is at least worth considering. It very well may be that the MoH is politically motivated of course, I can totally believe that, but the same applies for the regulator, in what is clearly a very politically contentious issue.

Expand full comment
Jun 13, 2022·edited Jun 13, 2022

I've recently been reading Freud/Jung for the first time. At the same time, I've been thinking about some of the questions from "Contra Dynomight on Sexy In-Laws", and I found an interesting confluence of ideas tying Jung's "Collective Unconscious" and "Archetypes" to Scott's question of different levels of biological drives.

From Scott's piece (https://astralcodexten.substack.com/p/contra-dynomight-on-sexy-in-laws):

> Since then, I’ve been thinking about this a lot. Is it correct to model myself as having (let’s call it) a reptile-level instinct of “genital friction feels good”, plus a separate mammal-level goal of “kids are cute”, while missing a human-level goal of “maximize inclusive genetic fitness, eg by donating to sperm banks”?

One of the key Jungian concepts is the "Collective Unconscious", the idea that all humans share a set of "archetypes" that are low-level concept clusters like "mother" or "father" (it's gendered in the Jungian model), and which are biologically determined, i.e. part of the shared human genetic heritage. These archetypes get projected into an individual's "personal unconscious" as "complexes" depending on their development/experiences, but Jung would say the archetypes themselves are universal. Jung takes this quite far and believes that there are lots of archetypes in the collective unconscious, including things that sound fuzzy to me like "wounded physician" and "hero on a quest", and that this explains the (claimed) shared symbology of dreams of all humans, but I think the more fundamental archetypes are particularly interesting as they shed light on Scott's question above.

The concept of archetypes gets picked up by the "Evolutionary Psychiatrists" and developed through the lens of Evolutionary Biology and ethology, with books like "Evolutionary Psychiatry" (Stevens, Price) that connects directly to Scott's framing:

> The triune brain provides a home for what we might call the 'triune mind'. Many thinkers, including Plato, St Augustine, La Rochefoucauld, Freud, and Jung, have observed that the mind seems to possess separate functional components which compete with one another for overall control of behaviour. Variously attributed to such organs as the 'head', the 'heart', and the 'bowels', reason, emotion, and instinct may display differing intentions when it comes to choosing a mate during courtship or displaying valour on the field of battle: 'Le coeur a ses raisons, que la raison ne connait point.' MacLean's anatomical studies give useful support to this long-standing concept of three minds in one (the neurological 'holy trinity').

> It begins to look, therefore, as if Jung was right when he guessed that the archetypal systems of the collective unconscious, if they could be given a local habitation and a name, must have their neuronal substrate in phylogenetically old parts of the brain.

There are some obvious problems here; aside from the "triune brain" concept seemingly having fallen out of favor, no actual evidence is produced for these archetypes aside from introspective/interpretive experience. Jungian analysis proposes that frustrated/suppressed archetypal expression results in some sort of psychic distress but I struggle to see how this theory is falsifiable. I suspect that the specific heteronormative claims of biologically-determined "mother" / "father" archetypes probably struggle to incorporate the evidence from modern attachment theory. And so on. However it's an interesting idea-cluster to put at one end of the spectrum; I think contrasting this with the Evolutionary Psychology perspective (Pinker, et. al.) could be fruitful.

Tying this all back to Scott's initial question, the "archetype" hypothesis is an expression of the idea that there is some inherited structure to cognition. The question is whether this inherited structure manifests as something vague like "wide behavior gradients" that provide a nudge to behaviors across a large area of the "behavior-space" but aren't resolvable as individual concepts, or at the other end of the spectrum, whether the inherited structure represents "narrow behavior gradients" that each strongly affect a small area of behavior-space and are potentially coherent as individual concepts. These could be "narrow low-level behavior gradients" that are more like Innate Releasing Mechanisms where it's a non-conceptual "stimulus => response" like "genital friction => pleasure/amorousness" and the behaviors are fairly tightly-specified (i.e. sex), or "narrow high-level behavior gradients" that represent actual concepts in some sense, i.e. pattern-matching on "my child => high-level conceptual desire to nurture/protect" (with the behavioral implementation of those high-level drive/concepts left to the individual's own learned behavior repertoire). The latter could perhaps be fruitfully mapped onto Jungian archetypes, inasmuch as such behavior clusters are present and universal.

Expand full comment

If anyone is interested in transgender issues, especially relating to women's sport, Sabine Hossenfelder made a really good video on the subject https://youtu.be/cZ9YAFYIBOU

(I am precommitting to not reply to any responses, cause culture wars, more like, culture snores)

Expand full comment

[Just posted this on the older open thread without noticing a newer one.]

Does anyone know of any data that gives an indication of AirBnB's sales?

Like, I know AirBnB won't directly give out that information (except during earning releases -- which are delayed). But I assume a "bot" could measure how many openings they have relative to a baseline or something. So I *assume* someone is publishing that data. I'm just unable to find it.

Expand full comment

That Brazilian guy sounds very dodgy. I can't tell if his "I did a PhD in 7 months!" is legit or not, the closest I can track down is a CV in Portuguese which translated to English says:

https://e4agencia.wixsite.com/lifestylesummit/dr-flavio-cadegiani

"Doctoral Student in Clinical Endocrinology at the Federal University of São Paulo / Escola Paulista de Medicina (Unifesp/EPM)"

That is "Student" not "Completed doctorate", unlike how his other degrees are listed. He may (or may not) have finessed "I did 7 months as a doctoral student then dropped out" into "I did a PhD in 7 months", I don't know, I can't track him down on the University of Sao Paulo website due to lacking the requisite skills. But if Marinos can get away with "for all we know", so can I:

"It’s probably the most awkward paragraph in a longer bio that—for all we know—could have been assembled from original Brazilian materials by a bored intern armed with Google Translate. Is it really fair to call that specific paragraph Cadegiani’s self-description?"

Well, I'm no bored intern and I had to rely on Google Translate since my Portuguese is not what it should be, but I'm betting this *is* Cadegiani's self-description.

"Getting three gold medals in three different national Science Olympiads—as Cadegiani claims —is incredibly unusual."

Precisely, which is why I'd like to see evidence of these claimed medals, claimed participation, etc. before I believe in Miracle Doctor's claims.

The rest of the links are very dodgy, too; that link to a fairy story about "so I can't tell you the guy's real name but trust me, he's rich and it's all true that the miracle doctor cured a second bout of Covid in 3 days, where the first time it took the hospital 30 days, and he did it with a cocktail of cheap generics" is a just-so story.

Cadegiani's 'big' breakthrough, once he threw over sports medicine and obesity treatments for Covid, seems to be claiming Covid is due to or affected by androgens (in men, at least) and using anti-androgens (proxalutamide) to treat it. This was the topic of a 2021 study (later retracted) and I'm not sure at all if there are any connections with the Chinese pharmaceutical company which published its own clinical trials in 2020 and are (or were) trying to licence it for treating Covid as well as prostate and breast cancer.

If I believe the BMJ article about Cadegiani et al getting their wrists smacked for their version of the clinical trial, it's pretty damn bad:

"The clinical trial of proxalutamide “disrespected almost the entire protocol” and may have contributed to the deaths of as many as 200 people, said the National Health Council (CNS), which oversees clinical research in Brazil.2 Some of those people were not adequately informed of the risks they were undertaking in the trial, and some did not know that they were taking part in one, it said.

...The treatment was prescribed by doctors as if it were an established medical treatment, said the CNS, although it was approved only for clinical studies. The number of people given the drug was also larger than the number approved for the trial, and they were administered through a private hospital network in the Amazon when the trial was approved in the capital, Brasilia."

So they threw a bunch of drugs at the patients, including ivermectin, and then claimed this was a miracle cure-all. I don't see why Marinos is so gung-ho for Cadegiani, save that he likes to throw ivermectin into the mix,and if Marinos believes ivermectin cures Covid on its own (and not as a secondary treatment for parasite infestation, relieving the burden on the immune system and general health of the patient), then any port will do in a storm.

Before reading that post, I knew nothing about Cadegiani. After reading it, I get strong alarm bells of "snakeoil salesman".

Expand full comment

Scott, if you’re listening, could you please reach out to me about my errant subscription gift?

Expand full comment

I often read in US reporting and elsewhere that the Sami are "the indigenous people of Sweden." But as far as I can tell, Swedes have been settled in Sweden for probably just as long. Does the distinction have some legitimate basis that I am missing?

Expand full comment

Would the person I gave the subscription to please reach out to me? Even if it’s only to say thanks.

Expand full comment

HELP. Somehow I inadvertently gave a gift subscription worth $100 to somebody in the last 40 minutes. Much as I love you guys I really didn’t mean to do that but I can’t find anywhere to turn to straighten it out other than going to my credit card company. Any advice?

Expand full comment

Anyone know how to get emails from Overcoming Bias? I dont have a RSS thingy.

Expand full comment

So, double-checking my awareness of the current state of algorithmic learning with people who know more about the field than I do.

People seem to have gotten okay-ish results in training algorithms to play the original Super Mario Bros (or at least the first couple of levels thereof). This seems to be mostly because that game has a couple of features they can take advantage of:

- It's easy to define "progress" (Mario's x-coordinate increasing), and it doesn't take much training data for the algorithm to encounter actions that affect progress (pressing the dpad right at any time).

- The basic game mechanics don't change a lot in later levels, and enemy movement patterns don't vary much. So early training data is mostly applicable to the whole game. (Although I've never seen an algorithm play a swimming level, or a castle level...)

- You don't really need to keep track of different game states like menu navigation, or otherwise keep track of how past actions have affected the current game state in hidden ways.

On the other hand, I don't think I've seen a report of anyone successfully getting an algorithm trained to play the original Legend of Zelda (or any other open-world type game, like Metroid, or I guess the Atari game "Adventure"). This seems to be a significantly more challenging task, because:

- Progress is much more nebulous and hard to discover. It takes a lot of gameplay before you encounter the main goal (defeating Ganon), or a sub-goal (finding a Triforce segment), or even a sub-sub-goal (finding any item). So it's easy for an algorithm to flail around for thousands of attempts without ever stumbling upon a "progress" counter, and there's no obvious incentive for it to take one path through the world over another.

- You regularly encounter new enemies with new movement patterns, and regularly discover new items which add actions you can take. So algorithms trained on early gameplay are likely to routinely hit walls any time a new element is introduced.

- You need to be able to switch items regularly, meaning you need to keep track of whether you're in the "gameplay" state or the "inventory menu" state. You also need to remember what items you've collected and change your strategy accordingly (e.g., water tiles are impassable initially, but passable after you collect the ladder).

So, am I correct in believing that a game like Zelda still poses a significant challenge to current algorithmic techniques, and no one in the field has really made much progress on that task so far? And, if so, would it be solvable in principle with the brute force "more data" approach, or is it an example of a task that might actually require a further paradigm shift in algorithmic learning?

I'm vaguely interested in poking at Zelda with the current publicly-available libraries to see how far I can train a bot to go, but I just wanted to check my thinking in principle before I put too much effort into it.

Expand full comment

Scott has had some recent interesting posts relating to Psychoanalysis. They also had those tantalizing ambiguous endings that signal that there is more to be explored…

Predict: will references to psychoanalysis, or even a more general interest writ large, become a larger part of Scott’s writing?

Reasons to think so:

—stuff about personal narrative, desires, etc. is important, and Scott seems to recognize that as such, and has been interested in such things thru his writing history

—in my opinion, the rationalist sphere is actually really quite uncompelling with regard to the above. As they are with metaethics. And maybe Scott, deep down, agrees with me.

Reasons to think against:

—on the whole, psychoanalysis obviously has a lot of nonsensical and pernicious stuff within its traditions as well

—why would Scott agree with you on this, B.B.?

—If technology continues to warp the 21st century at increasing rates, psychoanalytic questions may not be less relevant, but they will seem that way/be a lot more obscured. And nobody will really write on them anymore. (Maybe.)

Expand full comment

https://www.reddit.com/r/unsong/comments/vbdeg6/dalle_minis_rendition_of_noahs_whale/

Dalle mind's take on Noah's whale for the unsong readers here

Expand full comment

1) Data, the android, has just replaced a shard of his “brain” with a new updated hardware version. Because of redundancy in the system no memory is lost. He realises that he has replaced all the shards and all the brain hardware over the years. Is he still the same consciousness that initiated the replacement years ago?

2) Data is put offline, his software, and stored memories are saved somewhere else. The old brain is destroyed and the new brain is basically the same as if what happened in 1) happened all at once.

Is 2) the same as 1) in terms of continuity of consciousness?

Expand full comment

I was playing around with GPT-3 in some hazy combination of "trying to understand its limits" and "trying to understand if I could usefully use it as a GM tool in RPGs," and I ended up with this:

> Human: Who wrote "Everyone is John"?

> AI: The game "Everyone is John" was written by Greg Costikyan.

So GPT-3 and I are ENEMIES, NOW!!

Expand full comment

Assuming a few people here are psychologists and/or have good knowledge of psychometric testing, so just wondered if anyone knew (and could reference) the g-loading of different aspects of IQ tests?

Like what are the g-loadings of verbal reasoning tests, working memory tests, reaction time tests, matrix reasoning, etc?

Also in the literature is this odd phenomenon of autistics often scoring really well on the completely nonverbal Raven's progressive matrices test, but not so well on the more verbal Wecshler test ( a guy I spoke to at Mensa suggests a similar discrepancy with the Cattell B and Binet tests they use).

Does anyone have any ideas about why this happens?

Expand full comment

How does libertarianism deal with children? All the theory I've read takes the moral philosophy as being free exchange between consenting individuals, but a lot of these arguments work a lot better if you assume the contracting individuals are at least roughly equal in terms of abilities and power.

My question isn't just 'how does a philosophy based on free association deal with children who clearly didn't choose to be there', but also, how does the philosophy justify extreme differences in outcome that impact the children of said autonomous parties (assume, for the purpose of this argument, that your hypothetical interlocutor is fine with extreme differences of outcome for adult individuals, as long as that outcome came about via free exchange).

Expand full comment

If you like Scott's fiction, you might enjoy this story I wrote about Earth tourists visiting a D&D type fantasy world: https://vocal.media/fiction/the-thunder-valley-guided-dragon-slaying-experience - it does some light trope deconstruction and has jokes. This is for a contest with a top prize of $10k, so if you like the story please take a couple of minutes to leave a comment and click the heart button - that will help the judges notice it out of the 3000+ other entries. I promise to donate some of the prize money to top ranked EA causes if I win.

Expand full comment
Jun 13, 2022·edited Jun 13, 2022

Any thoughts on the guy at google saying their AI is sentiment?

https://archive.ph/1jdOO

Edit: see “philosophy bear”’s post below

Expand full comment

The categories were made for man, not man for the categories.

https://slatestarcodex.com/2014/11/21/the-categories-were-made-for-man-not-man-for-the-categories/

Now in California, bees are categorized as fish (under the California Endangered Species Act):

https://www.courts.ca.gov/opinions/documents/C093542.PDF summarized at https://www.fieldandstream.com/conservation/california-bumble-bees-are-fish/

Expand full comment

Has there been any work on Robin Hanson's CEO Prediction Market idea? [1] With all the attention prediction markets are getting, I'm surprised it hasn't happened. Similarly, you could make markets for GDP or the stock market that were conditional on a presidential election result or specific congressional bill, though I'd be worried about market manipulation.

[1] https://www.overcomingbias.com/2008/04/if-i-had-a-mill.html

Expand full comment

Last year I applied for a job in a technical role at a very big company. It had all the typical hiring stages, except one that I've never seen or heard about before or since: the 2nd stage (after the IQ-ish test) was a series of interview question videos, where you watch someone virtually ask you an interview question then you have to answer into your webcam and mic, with your answer recorded. Like a conversation, but their half is pre-recorded. At the end, it collates all your speech (and face movement??) data, and sends you back a report with something like:

"Our systems detect that you are a HIGHLY TECHNICAL and SYSTEMS-LEVEL thinker who prefers to work ALONE OR IN SMALL GROUPS with DATA or ENGINEERING PROBLEMS." etc. etc.

15 mins later, I got an automated email saying I didn't get the job. NB I did try not to sound like a nerd.

Not getting that job wasn't an issue in the scheme of things, but I can't stop thinking about that test. Was it scraping my speech for buzzwords and checking them off against a list of "things the candidate must say" ? Or was it looking at the overall landscape of my responses like the phrasing, vocab, and stress/emphases? If I ever have to do one of these again, should I try to game it somehow by optimising for buzzword density? Anyone have any insight here?

Expand full comment
Jun 13, 2022·edited Jun 13, 2022

There are newtonian fluids: if you you poke them hard with a stick, they splash and the stick gets wet.

There are non-newtonian fluids: if you hit them with a stick, the stick bounces off.

And there are anti-inductive fluids: if you hit them with a stick, anything can happen, including the fluid grabbing the stick and hitting you back.

Expand full comment

Slight update on my predictions on the outcome of Ruso-Ukrainian war. Previous prediction is here: https://astralcodexten.substack.com/p/open-thread-224/comment/6591197?s=r.

10 % on unambiguous Ukrainian victory (unchanged).

Ukrainian victory is defined as either a) Ukrainian government gaining control of the territory it had not controlled before February 24, regardless of whether it is now directly controlled by Russia (Crimea), or by its proxies (Donetsk and Luhansk "republics”), without losing any similarly important territory and without conceding that it will stop its attempts to join EU or NATO, b) Ukrainian government getting official ok from Russia to join EU or NATO without conceding any territory and without losing de facto control of any territory it had controlled before February 24, or c) return to exact prewar status quo ante.

28 % on compromise solution which both sides might plausibly claim as victory (down from 30 % on May 15).

62 % on unambigous Russian victory (up from 60 % on May 15).

Russian victory is defined as Russia getting something it wants from Ukraine without giving any substantial concessions. Russia wants either a) Ukraine to stop claiming at least some of the territories that were before war claimed by Ukraine but de facto controlled by Russia or its proxies, or b) Russia or its proxies (old or new) to get more Ukrainian territory, de facto recognized by Ukraine in something resembling Minsk ceasefire(s)* or c) some form of guarantee that Ukraine will became neutral, which includes but is not limited to Ukraine not joining NATO. E.g. if Ukraine agrees to stay out of NATO without any other concessions to Russia, but gets mutual defense treaty with Poland and Turkey, that does NOT count as Russian victory.

Reasons:

Just like last update was because of German elections, this one is because of French elections, where after the first round it looks fairly plausible that Macron will lose his parliamentary majority (fairly unprecendented and thus surprising). And if it happens, he will lose it parties less favorably disposed to Ukraine than him.

French elections to national assembly are more important for French foreign policy than German election in North Rhine-Westphalia for German foreign policy, but, Germany is now important for future direction of the conflict than France, so I guess they count roughly the same.

If he retains his majority and nothing major happens, I am going to update bach to 60/30 split in favor of Russian victory, I am going to update to 65/25 in favor of Russian victory.

Expand full comment

In my discussions a couple weeks ago on functionalism vs. physicalism in consciousness, i ended up trying to argue that information is not physical. Leaving consciousness aside for the moment, I now summon prof. Robert Alicki on my side who in the first chapter makes my point way better that I ever will:

https://arxiv.org/abs/1402.2414

In fact, i am now not even sure if I formulated my position on the nature of information independently or i read this paper, took from it and promptly forgot to ever having read it

Expand full comment

The study linked to by Crimson Wool seems tailor-made to maximise the Lizardman constant to me, at least in the weak form of "Obama is the anti-Christ". I may not actually be prepared to do violence against my political enemies, but I'm sure going to express my opposition to them in the strongest terms you offer me.

Expand full comment

Reflecting on the Blake Lemoine situation, how likely is it that when the first sentient AI does arise it will be dismissed as "just faking it", and what sort of evidence could be gathered to the contrary? Here there were some obvious tells (e.g. the model talking of "meditating", despite it just doing feedforward I/O), but presumably future models will be smart enough to avoid such traps.

Expand full comment

"only a 20% chance we can get it without something symbol-manipulation-y in particular (compared to my 66%)"

Note the distinction between we *will* get it vs. we *can* get it without something explicitly symbolic. There's a real chance that some transformer with a symbolic module bolted on could get us to AGI faster than without, but that doesn't mean we'll have *needed* it to get there. And we might very well have gotten there a year or two later without it.

Expand full comment
Jun 13, 2022·edited Jun 13, 2022

I think the "which party got more extreme?" was kind of a shit show in part because Scott Alexander hasn't weighed in on any of the controversies.

Someone saying in the comments, "How can you not see that the election was obviously stolen?" is a dead end because we, as random commenters, are not going to be able to convince each other of anything.

But we are all here and read Scott Alexander because of the exhaustive analysis and interesting thought. If this blog had a post about the 2020 election or banning abortion that would be very illuminating.

That was sort of the problem with the which party got more extreme post. "Can't you see that Xxx is obviously nuts?" is a common sentiment. Taking a rationalist deep dive on some of these "obviously nuts" issues might actually change some minds and lead to a real dialogue.

Expand full comment

So just today there were quite a few "innovative" strategy game announcements. Victoria 3 announced a 2022 launch and had a gameplay trailer, Great Houses Of Calderia had a trailer and I think an EA release date? Grey Eminence didn't do anything, shoulda got on the hype train.

Finally Bethesday announced ARA: Untold History, a new strategy game, Civ+++ based on screenshots hidden on the site, since the trailer was just animation and a voice over. Weird name choice. Ara ara, uwu, so many arms on that human body study Mr. Davinci. Perhaps a secret hentai easter egg? Probably not.

Also Hooded Horse still on a tactical rpg, survival sim, general strategy publishing spree.

Given this, plus some other less relevant strategy news:

For people who play any kind of strategy or simulation game from Three Houses to CK3 to potentially Vicky3, or GalCiv4 or w/e, what kind of stuff are you wishing would get into this relative wealth of strategy games the last few years?

More character stuff like Stellar Monarch, Star Dynasties, or Alliance Of The Sacred Suns? Or I guess Old World? More logistical stuff and simulation like Shadow Empires or maybe Distant Worlds 2? Economic sim and at least a facade of diplomacy ala Grey Eminence and Vicky 3? CK3 style memes?

Iconic EU4 mod Anbennar just had a massive update. So more fantasy grand strategy? Potential magical stuff? The Haless or w/e update for the eastern part of the map added a lot of fancy, and impressive given EU4 lacks the updated modding options, magical stuff.

We still don't have any amazing intrigue or espionage. Though Espiocracy, another Hooded Horse acquisition, is going there at least for the modern age historical genre. Or maybe Terminal Conflict? The Cold War leaves me cold but the blue/orange 80s/90s "terminal" UI from which the game takes a name does look cool and flavorful.

Expand full comment

Informal poll: I think Star Wars is not actually science fiction, largely because it has no science. Rather, I think it's a traditional fantasy "farm boy meets a wise old man, gets a magic sword, goes on an adventure" type story that just happens to have space trappings. Luke never runs an experiment, he puts himself in the moral mindset some old guys tell him to take and this makes him better at swinging a sword and jumping and flying spaceships. I think, instead, Star Wars is one of the quintessential works of Space Operas; works with fantasy narrative structures that happen to be set in space. (Even hotter take: by this definition Mistborn is closer to science fiction than Star Wars is!)

But even a number of my friends who read and watch a lot of science fiction and fantasy don't draw this distinction. Is distinguishing Space Opera (and Cyberpunk and all the other "sci fi setting" genres) from Science Fiction correct? Important? Can you name categories you think this is more or less important than to help me calibrate?

Expand full comment

A question on different brands of the same drug. In short, is there / does it seem reasonable that there might be a difference in effect between the Brown & Burk version of Gabapentin, and the (generic?) version available in Canada?

My Grandma recently had to use the Brown & Burk version, and found it to work much better for her neuropathy. Same dosage and everything.

Some more detail: she was on a cruise and forgot her regular supply, and picked up what was available on the cruise, the Brown & Burk. Instead of the pain being managed, the pain was effectively gone. The switch back occured several days after the cruise, when the new supply ran out and she went back to regular. Thus, it sounds like it was probably mostly the different Gabapentin, not something else (changes to diet or environment on the cruise, e.g.)

So, 2 questions: does this sound right, that different brands of this drug could have substantially different efficiacies? And if so, does anyone know if there's a good way to get it in Canada? She's close enough to the US border that she is willing to go down to pick something up.

Expand full comment

All of the popular science articles on the web say that women are at an increased risk of many cancers when they drink any amount of alcohol. Is this pop consensus aligned with the most careful research? Can anyone point me in the direction of the best resource for information on alcohol effects on women? Thanks!

Expand full comment

Speaking of your mistakes page, you conclude the second paragraph there by saying "in reverse chronological order, starting with my old blog" -- surely you mean "in reverse chronological order, including my old blog", since you're starting with the most recent things on the current blog.

Expand full comment

I wanted to share my article on claim's of LaMDA' sentience, see if anyone has any thoughts:

https://philosophybear.substack.com/p/regarding-blake-lemoines-claim-that?s=w

Expand full comment

I’ve been thinking more about the “extreme” and “x-wing” parties posts, and I am becoming much more skeptical that this is being approached the right way. Most of the evidence for one point or another comes from polls asking people how they feel on something, which only covers stated, not revealed, preferences. So I don’t think this could distinguish the case of “a party is becoming more extreme” from “a party’s base is becoming more likely to signal their loyalty to the tribe by taking an extreme position.” These may be correlated, but then again, they may not, and IDK how to separate them.

Expand full comment

It was genuinely fun while it lasted, and would be happy to do it again some time. (also, I appreciate the edit.)

Expand full comment
User was banned for this comment. Show
Expand full comment
Comment deleted
Expand full comment
Comment deleted
Expand full comment
Comment deleted
Expand full comment
Comment deleted
Expand full comment
deletedJun 13, 2022·edited Jun 13, 2022
Comment deleted
Expand full comment
Comment deleted
Expand full comment