Astral Codex Ten Podcast

Latest Episodes:

The Obligatory GPT-3 Post Jun 12, 2020

Show notes

https://slatestarcodex.com/2020/06/10/the-obligatory-gpt-3-post/

I would be failing my brand if I didn't write something about GPT-3, but I'm not an expert and discussion is still in its early stages. Consider this a summary of some of the interesting questions I've heard posed elsewhere, especially comments by gwern and nostalgebraist. Both of them are smart people who I broadly trust on AI issues, and both have done great work with GPT-2. Gwern has gotten it to write poetry, compose music, and even sort of play some chess; nostalgebraist has created nostalgebraist-autoresponder (a Tumblr written by GPT-2 trained on nostalgebraist's own Tumblr output). Both of them disagree pretty strongly on the implications of GPT-3. I don't know enough to resolve that disagreement, so this will be a kind of incoherent post, and hopefully stimulate some more productive comments. So:

OpenAI has released a new paper, Language Models Are Few-Shot Learners, introducing GPT-3, the successor to the wildly-successful language-processing AI GPT-2.

GPT-3 doesn't have any revolutionary new advances over its predecessor. It's just much bigger. GPT-2 had 1.5 billion parameters. GPT-3 has 175 billion. The researchers involved are very open about how it's the same thing but bigger. Their research goal was to test how GPT-like neural networks scale.

Before we get into the weeds, let's get a quick gestalt impression of how GPT-3 does compared to GPT-2.

Here's a sample of GPT-2 trying to write an article:

Take the New Nootropics Survey Jun 08, 2020

Show notes

A few years ago I surveyed nootropics users about their experiences with different substances and posted the results here. Since then lots of new nootropics have come out, so I'm doing it again. If you have nootropics experience, please take The 2020 SSC Nootropics Survey. Expected completion time is ~15 minutes.

Thanks!

Problems With Paywalls Jun 06, 2020

Show notes

https://slatestarcodex.com/2020/06/04/problems-with-paywalls/

I hate paywalls on articles. Absolutely hate them.

A standard pro-business argument: businesses can either make your life better (by providing deals you like) or keep your life the same (by providing deals you don't like, which you don't take). They can't really make your life worse. There are some exceptions, like if they outcompete and destroy another business you liked better, or if they have some kind of externalities, or if they lobby the government to do something bad. But in general, if you're angry at a business, you need to explain how one of these unusual conditions applies. Otherwise they're just "helping you less than you wish they did", not hurting you.

And so the standard justification for paywalls. Journalists are providing you a deal: you may read their articles in exchange for money. You are not entitled to their product without paying them money. They need to earn a living just like everyone else. So you can either accept their deal – pay money for the articles – or refuse their deal – and so be left no worse off than if they didn't exist.

But I notice feeling like this isn't true. I think I would be happier in a world where major newspapers ceased to exist, compared to the world where they exist but their articles are paywalled. Take a second and check if you feel the same way. If so, what could be going on?

Book Review: Origin of Consciousness in the Breakdown of the Bicameral Mind Jun 03, 2020

Show notes

https://slatestarcodex.com/2020/06/01/book-review-origin-of-consciousness-in-the-breakdown-of-the-bicameral-mind/

Julian Jaynes' The Origin Of Consciousness In The Breakdown Of The Bicameral Mind is a brilliant book, with only two minor flaws. First, that it purports to explains the origin of consciousness. And second, that it posits a breakdown of the bicameral mind. I think it's possible to route around these flaws while keeping the thesis otherwise intact. So I'm going to start by reviewing a slightly different book, the one Jaynes should have written. Then I'll talk about the more dubious one he actually wrote.

My hypothetical Jaynes 2.0 is a book about theory-of-mind. Theory-of-mind is our intuitive model of how the mind works. It has no relation to intellectual theories about how the mind is made of cognitive algorithms or instantiated on neurons in the brain. Every schoolchild has theory-of-mind. It goes like this: the mind is an imaginary space containing things like thoughts, emotions, and desires. I have mine and you have yours. I can see what's inside my mind, but not what's inside your mind, and vice versa. I mostly choose the things that are in my mind at any given time: I will thoughts to happen, and they happen; I will myself to make a decision, and it gets made. This needs a resource called willpower; if I don't have enough willpower, sometimes the things that happen in my mind aren't the ones I want. When important things happen, sometimes my mind gets strong emotions; this is natural, but I need to use lots of willpower to make sure I don't get overwhelmed by them and make bad decisions.

Bush Did North Dakota May 31, 2020

Show notes

https://slatestarcodex.com/2020/05/28/bush-did-north-dakota/

Continuing yesterday's discussion of fake news:

Guess et al says that 46% percent of Trump voters endorsed the Pizzagate conspiracy theory. Does this mean fake news is very powerful?

We can compare this to belief in various other conspiracy theories, as measured by the 2016 Chapman University Survey Of American Fears. About 24% believe there's a government conspiracy to cover up the truth about the moon landing, 30% about Obama's birth certificate, and 33% about the North Dakota crash.

This last one is especially interesting because there was no unusual crash in North Dakota when the survey was written. The researchers included it as a placebo option to see if people would endorse a conspiracy theory that didn't exist. 33% of them did.

Before we make fun of these people, consider: there's a strong presumption that surveys don't contain made-up questions. There was no "don't know" option included on the poll, just various shades of "agree" or "disagree". In order to condemn the people who "agreed" that the government was probably covering up the crash, we would have to assert that the more correct answer was "disagree". In other words, that people should have an assumption of trusting the government, until they get some specific reason to distrust it. You can make that argument, but it's not obvious. You could also start from the opposite assumption, where the government is guilty until proven innocent.

[Classic] Guided by the Beauty of Our Weapons May 30, 2020

Show notes

https://slatestarcodex.com/2017/03/24/guided-by-the-beauty-of-our-weapons/

[Content note: kind of talking around Trump supporters and similar groups as if they're not there.]

Tim Harford writes The Problem With Facts, which uses Brexit and Trump as jumping-off points to argue that people are mostly impervious to facts and resistant to logic:

All this adds up to a depressing picture for those of us who aren't ready to live in a post-truth world. Facts, it seems, are toothless. Trying to refute a bold, memorable lie with a fiddly set of facts can often serve to reinforce the myth. Important truths are often stale and dull, and it is easy to manufacture new, more engaging claims. And giving people more facts can backfire, as those facts provoke a defensive reaction in someone who badly wants to stick to their existing world view. "This is dark stuff," says Reifler. "We're in a pretty scary and dark time."

He admits he has no easy answers, but cites some studies showing that "scientific curiosity" seems to help people become interested in facts again. He thinks maybe we can inspire scientific curiosity by linking scientific truths to human interest stories, by weaving compelling narratives, and by finding "a Carl Sagan or David Attenborough of social science".

Creationism, Unchallenged May 29, 2020

Show notes

https://slatestarcodex.com/2020/05/28/creationism-unchallenged/

How much should responsible news organizations report on stupid things?

If they don't report at all, the stupid things go unchallenged. But if they report too much, then they signal-boost the stupid thing and give it free publicity (eg Donald Trump). Also, people who mistrust the media might reflexively support the stupid thing just because the media hates it (eg Donald Trump). Also, the more time you waste covering stupid things, the less time you have for real news (eg Donald Trump).

I recently read Causes And Consequences Of Mainstream Media Dissemination Of Fake News: Literature Review And Synthesis, which argues that the news might be covering too many stupid things right now. The authors note that "only 2.6% of visits to current affairs articles were to fake news websites" (though other sources suggest more) and that the mainstream press bears some responsibility for spreading inaccuracies beyond this small demographic. But they also understandably worry that maybe if the mainstream press wasn't so aggressive in covering and debunking fake news, then fake news would go uncorrected.

"My Immortal" As Alchemical Allegory May 28, 2020

Show notes

https://slatestarcodex.com/2020/05/26/my-immortal-as-alchemical-allegory/

From Vox: Solving The Mystery Of The Internet's Most Beloved And Notorious Fanfic. The fanfic is "My Immortal", a Harry Potter story so famous that it has its own Wikipedia page, and articles about it in Slate, Buzzfeed, and The Guardian.

It's famous for being really, really bad. Spectacularly bad. Worse than it should be possible for anything to be. You wouldn't think you could get The Guardian to write an article about how bad your fanfiction was, but here we are. Everyone agrees that it must have taken a genius to make something so awful, but until recently nobody knew who had authored the pseudonymous work. The Vox article investigates and finds it was probably small-time author Theresa Christodoupolos, who goes by the pen name Rose Christo.

But this leaves other mysteries unresolved. Like: what is going on with it? Its plot makes little sense – characters appear, disappear, change names, and merge into one another with no particular pattern. Even its language is fluid, somewhere between misspelled English and a gibberish that can at best produce associations suggestive of English words.

Coronalinks 5/18/20: When All You Have Is a Hammer, Everything Starts Looking Like a Dance May 20, 2020

Show notes

https://slatestarcodex.com/2020/05/18/coronalinks-5-18-20-when-all-you-have-is-a-hammer-everything-starts-looking-like-a-dance/

It is the sixty-first day of shelter-in-place. Anti-lockdown protesters have stormed your state capitol, chanting Nazi, Communist, ISIS, and pro-Jeffrey Epstein slogans to help you figure out they're the bad guys. Inside, the Governor has just finished announcing his 37 step plan to reopen the state over the next ten years. You kind of feel like he should be a little more proactive, but the protesters outside have just unfurled a Khmer Rouge flag, so you hold your tongue.

Meanwhile, a band of renegade economists, tech billionaires, and MIT professors has just announced a bold disruptive Manhattan-Project-style moonshot: send a team of researchers to the swamps of Florida, where legends speak of a Fountain of Youth whose water can cure any malady. But disaster strikes when Florida's governor announces that exploration is not an essential activity, and threatens to release the quarantine enforcement lions. The nation looks to the White House to solve the growing conflict, but President Trump is too busy evangelizing his latest coronavirus cure: eating those little packets of silica gel in food that say DO NOT EAT. As the Western States Pact and the Eastern Bloc inch closer to war, all that the rest of us can do is strive to stay as well-informed as possible, trying to make sense out of an increasingly nonsensical situation. So:

[Classic] Cardiologists and Chinese Robbers May 16, 2020

Show notes

https://slatestarcodex.com/2015/09/16/cardiologists-and-chinese-robbers/

It takes a special sort of person to be a cardiologist. This is not always a good thing.

You may have read about one or another of the "cardiologist caught falsifying test results and performing dangerous unnecessary surgeries to make more money" stories, but you might not have realized just how common it really is. Maryland cardiologist performs over 500 dangerous unnecessary surgeries to make money. Unrelated Maryland cardiologist performs another 25 in a separate incident. California cardiologist does "several hundred" dangerous unnecessary surgeries and gets raided by the FBI. Philadelphia cardiologist, same. North Carolina cardiologist, same. 11 Kentucky cardiologists, same. Actually just a couple of miles from my own hospital, a Michigan cardiologist was found to have done $4 million worth of the same. Etc, etc, etc.

My point is not just about the number of cardiologists who perform dangerous unnecessary surgeries for a quick buck. It's not even just about the cardiology insurance fraud, cardiology kickback schemes, or cardiology research data falsification conspiracies. That could all just be attributed to some distorted incentives in cardiology as a field. My point is that it takes a special sort of person to be a cardiologist.

Our TOPPODCAST Picks

Follow Us

Stay Connected

Related Podcasts

Links

Stay Connected