croissants

I went to a talk by an author of one of these papers (David Rand). IIRC the format of the paper's experiment was collecting people with conspiracy theory beliefs and having LLMs with different strategies try to reduce their confidence in the conspiracy theory.

The best strategy, perhaps reassuringly, was tailoring facts to the person's worldview. I don't mean massaging facts, I mean that if the person thinks 9/11 was an inside job because of something about melting steel, then the LLM will present some sourced facts about how that actually works. And then it will do the same thing for all six of the sub-beliefs that support their overall belief. This was key to the strategy: the LLM has enough breadth and recall to exactly and patiently rebut most mistakes a person is making without making them defensive.

This strategy handily beat other things like flattery, selectively lying, etc. So while LLMs can clearly persuade, the evidence suggests it is easiest for them to persuade belief in something that is actually true.

This was in 2024 or 2025. Things might be different now.

The past few months have, among other things, illustrated to me how many people feel hurt by past experiences with mathematicians, and how this, not always unreasonably, colors their opinions of the field and its future today.

There are much easier and more prestigious ways of fulfilling university service requirements than reviewing and editing for TMLR, so I think it is correct in spirit to label it as volunteering.

I took a course on Freud in college, and what struck me at the time was that the framework of psychoanalysis seemed basically cruel, in that the analyst was free to say that either acceptance or rejection of their claims was evidence for them, and that it would be maddening to deal with that for any extended period of time.

The paper does more than that. It measures wet bulb temperatures at different stops with different combinations and varieties of trees and bus stops, provides decent evidence of variation among these combinations, and attempts to explain the variations. For example, the roles played by evaporation from leaves and shelter material weren't obvious to me before reading.

More generally, this method of interacting with the most simplified interpretation of science and then criticizing its simplicity isn't useful to the criticizer or the criticized.

+1 for Consumer Reports. They're not expensive either, something like $5 per month. If they keep you from buying a bad fridge, it pays for itself!

Their recentish coverage of lead in foods is a bit embarrassing though, since they used a California standard for dosage limits that even the EU would blush at.

I chalk that up as lame but kind of inevitable mild corruption that doesn't really go anywhere -- Hunter seems more like an embarrassing addict child that a caring parent has no choice but to put up with than somebody who can truly peddle influence over his father -- and feel that this administration is a vast difference in degree of corruption. Degrees matter!

Yeah, The Economist regularly writes these articles that spend half the time praising the intense competition in certain sectors of the economy and half the time bemoaning the influence of state-owned enterprises (or other CCP instruments) in others. It seems to vary a lot by specific industry.

At least in moderately urban parks in the USA, one complicating factor is that a lot of the dirt has dog poop scattered on top.

I'm surprised that nobody has mentioned Blindsight. I don't think it's a spoiler to say that it is a book about the place of human intelligence in a universe with other options, both biological and artificial.

I don't think there's a good discussion to be had about this article. The referenced paper is based on two data sources. One is a longitudinal study (great!) of "hundreds" of children in one region (not bad!) and only checks parenthood at the rather young age of 23-24 (uhhh). The other study is based on Mechanical Turk questions. In both the article and the fullest version of the paper I can find online [1], no further details are provided -- not the size of the effects, not the statistical power, nothing.

[1] https://psycnet.apa.org/doiLanding?doi=10.1037%2Febs0000374

Please elaborate?

I'm not sure what word to apply to facing three familial tragedies, any one of which might seem like more than a life's fair share of misfortune, with that kind of perseverance. Inspiring is too glib, inhuman is too alienating. Whatever kind of mettle that is, I hope to never have to prove it.

Yeah, I think this is one of those domains where frictions (or lack thereof) matter. It is much faster, cheaper, and easier to send out a significant volume of material on the Internet than through the mail. In theory, somebody could send a bunch of obscene trolling letters, but having to lick all those envelopes seems to deter most of the people who'd be otherwise tempted.

(To appear at STOC 2025 [1], one of the top CS theory conferences, so it's passed peer review.)

[1] https://acm-stoc.org/stoc2025/accepted-papers.html

> they can be applied to many different situations, even business management and interpersonal relationships

Disclaimer: most people do not think "Machiavellian" is a flattering descriptor

> In 4 years of working cases I would estimate 1 in 6 jurors are above the 85-115 IQ range of average intelligence, and maybe half are at or below the 100 line.

Maybe I'm missing the joke, but isn't IQ meant to follow a normal distribution with a mean/median of 100 with a standard deviation of 15, in which case you'd expect half of jurors to be below 100 and ~15% to be above 115, which is pretty close to what you've seen?

I think a modern USA 1 year old has about a 99.97% chance of making it to adulthood. That means that if a modern USA adult loses a young child, there's a decent chance they don't know anybody who has had that experience.

The ancient (and even, as you point out, very slightly pre-modern) world had a lot of "infrastructure" in place to deal with this, there were rituals and ceremonies and familiar people who knew what you were going through, and most of that is gone now.

It is indeed an enormous change.

I was going to ask about this number, because it seems high enough to be statistically improbable, but back-of-the-envelope arithmetic says otherwise: there are about 10 cases of pancreatic cancer per 100,000 people per year [1], so let's say each person has a 1 in 10,000 chance of a diagnosis each year. If you know somebody for 50 years, there's a 1 in 200 chance they receive a diagnosis in that time, so you'd expect to need to know 2000 people to eventually know 10 diagnosed people. 2000 is a lot, but "knowing" a person is a pretty loose term, and pancreatic cancer has a miserably high death rate within 5 years, so it's unfortunately plausible.

[1] https://seer.cancer.gov/statfacts/html/pancreas.html

> If the PhD is losing its lustre, it’s because the Universities took the shine off

Also, circulating particularly weird dissertations for the express purpose of angering people has gotten a lot more rewarding

If you want the same kind of thing with no useful information, how about this report on the accuracy of a sign? https://soranews24.com/2024/06/27/is-this-7-eleven-sign-in-j...

Sure, so it seems fair for him to offer advice on how to apply AI or start a company based on it, but doing AI research means generating new knowledge about AI itself, and I don't see any evidence of that.

Previous discussion: https://news.ycombinator.com/item?id=22101066

IMO the confident Tweets about how to become a good researcher and hire good researchers look pretty weird next to the lack of any apparent research papers (or even visible research products) five years later.

> totally independent

How do you arrive at this conclusion? Individuals don't have to tell you where their money comes from. They might even be easier to influence/buy than the people inside the big news institutions.

The book also has a sex scene featuring (spoiler, I guess) a 15-year-old girl. It's probably possible to write something like that in a way that isn't gross, but Stephenson definitely didn't pull it off.

I like Stephenson, he's written some great stuff, but even ignoring that scene, Snow Crash is IMO just a bad book to read once you're out of high school.

Do you want the version without data collection/sharing to still be free?

At least in computer science, postdocs have other benefits:

1) they can be well paid, like high five to very low six figures;

2) they can be an extra year or two to figure out your own research direction with some help but not much oversight;

3) they can be much easier to get than industry positions -- sometimes requiring little more than a solid publication record and advisor recommendation;

4) if you've been in a university environment for ~a decades and liked it, it might strike you as an easy path to keep doing that (this is probably the worst reason, though).

This is skewed by computer science postdocs at highly ranked schools, though. Yes, people taking these positions face opportunity costs, but the actual experience can be pretty nice.

A possible line of reasoning is that drugs should be legal, but the property and violent crimes committed around them shouldn't be, in the same way that adults are legally permitted to drink alcohol, but they're not legally permitted to drive drunk. The "ruining cities" is about the crimes, not the drugs themselves.

(I think.)

Pat and cynical oversimplifications are bad for discourse, because they suggest that a default angry response is correct and, coincidentally, frees you from having to think harder about anything.

Don't give in!

My fairly generic PhD advice, as somebody who did one, graduated a bit early, and is happy about the process and where it's taken me:

* Choose your advisor with care. This is not very easy as an applicant looking at professors' websites, but if you are admitted, any good school will probably have an in-person or virtual admitted students day where you can talk to current students out of faculty earshot. Take advantage of these times to ask about your potential advisor. A truly bad advisor will probably produce at least one person who will warn you about them. If you can't do this in person, try to get a quick phone/video chat -- something off-record where they can be honest. I was always happy to do these for my advisor, because I liked him and wanted him to get more good students. Conversely, I know people who were warned off specific advisors during these events, for good reasons. A bit of subjectivity: a good advisor at a decent school is usually better than a bad advisor at a good school.

* The financial niceness of doing a PhD in field X seems to correlate pretty well with the current job market value of a masters in field X, at least partially for reasons of leverage -- if you can leave and transition into a cushier job, advisors have to provide a bit more value. Computer science scores highly on this metric.

* There is a ton of negativity about PhDs in places like HN. This isn't unjustified: doing a PhD with a bad advisor can be a very bad experience. At the same time, I think "person who had a bad PhD experience" is also "person who writes comments on the internet" with higher probability than "person who had a good PhD experience".