Tyler Cowen

GPT showing off on Conversations with Tyler

Joy2 Comments

Since posting it in January of 2023, our post on a failure of ChatGPT has been one of the most popular articles of the year in terms of web traffic. Read ChatGPT Cites Economics Papers That Do Not Exist

LLMs will make up nonsense lies, but it might sound sophisticated and serious. As human editors, we are not used to detecting that kind of problem, because usually if something sounds serious that means it has been written by a serious thoughtful person (usually).

That’s a problem, but ChatGPT is really good at some tasks.

Instead of pointing out the fact that ChatGPT is not perfect, Tyler Cowen wants people to see the potential. He has done an entire podcast interview with ChatGPT, asking it to take on the voice and role of deceased writer Jonathan Swift.

I’ll share part of the transcript:

COWEN: In 1709, you, Jonathan Swift, still were calling yourself a Whig. What exactly induced you to switch to the Tory side of the ledger?

SWIFT: My shift from the Whigs to the Tories was a gradual and complex process, influenced by several factors, including personal connections, political developments, and ideological disagreements.

During my early years in politics, I was indeed more aligned with the Whigs, who championed principles such as constitutional government, religious toleration, and the supremacy of Parliament over the monarchy. However, over time, I became increasingly disillusioned with the Whigs, particularly as their internal factions and power struggles intensified.

The Whigs’ support for the war with France, which began in 1702, was a point of contention for me. I believed that the war was costly…

The replies seem nuanced and insightful. Nothing made me suspect that they were factually inaccurate, although I do not know enough about Irish history to judge.

Is there any human who could have produced this script? I think so, although it would have required a lot of work. If one of these replies is better than anything a human Swift scholar would produce, how would we know?

GPT4 can write good summaries for the work of a prolific author like Swift, because the model can train on lots of examples.

GPT4 could probably write a good biography of a modern figure by pulling together all of the writing by them and about them. Maybe GPT4 could efficiently scrape up all mentions of this figure online and synthesize them faster than a human scholar. However, we observed GPT3 completely making up citations when we tried to get it to do economics summaries.

I’m concerned that humans will use GPT4 to write but not do the requisite fact-checking. That could introduce a new corpus of work that the next LLMs will train on, which might be full of lies. Humans might not admit to using GPT, and therefore we wouldn’t have a mechanism for using extra scrutiny on AI-generated writing from 2023. Humans can make mistakes too… so the ultimate solution could be an all-powerful AI that somehow does begin with a fairly accurate map of the world and goes around fact-checking everything faster than human editors ever could.

Always Be Posting

James Bailey4 Comments

The key lesson, the thing I would impart to any aspiring bloggers, content creators, or newsletter proprietors, is that the cornerstone of internet success is not intelligence or novelty or outrageousness or even speed, but regularity. There are all kinds of things you can do to develop and retain an audience — break news, loudly talk about your own independence, make your Twitter avatar a photo of a cute girl — but the single most important thing you can do is post regularly and never stop.

Granted, “work hard and keep doing the thing you’re doing” is probably above-replacement advice for any kind of entrepreneurial activity. But it’s particularly true for online content creation. As Yglesias suggests, the internet and feed-based social platforms have constructed an insatiable demand for content, so if you can produce content mechanically, without requiring expensive resources (such as time, wit, or subject-specific knowledge), you’re in an excellent position to take advantage. But most importantly, this demand is so insatiable that there is currently no real economic punishment for content overproduction. You will almost never lose money, followers, attention, or reach simply from posting too much.

That is from an excellent post by Max Read. It is fairly short and has some good examples, so I recommend reading the whole thing. Tyler Cowen made a similar point back in 2019:

There’s a certain way in which on the internet you can’t be overexposed. There’s just a steady stream of you, it feels like being overexposed compared to the standards of 1987 or whenever, but in fact it’s not and people are picking and choosing. And you end up just dominating a particular space in a particular kind of way. And I think most older people have not made that transition mentally to understanding how you should exist intellectually on the internet.

Of course, this is also part of Joy’s idea with this site and why we write every day:

It’s just time to start writing more. This is a model that I have learned from Tyler Cowen, and most writers I admire write every day whether or not they have time for it. David Perell has tweeted that writing and thinking are the same thing. Thus, if you are a thinker, writing is not a waste of time. Writing is the thing you are doing anyway in your muddled head.

Thus: Always Be Posting.

Gambler Ruined: Sam Bankman-Fried’s Bizarre Notions of Risk and the Blow-Up of FTX

Scott Buchanan3 Comments

The drama continues for Sam Bankman-Fried (SBF), the former head of now-bankrupt crypto exchange FTX. This past week has been giving a series of interviews, in which he (the brilliant master, the White Knight, of the crypto world a mere month ago) is trying to convince us (potential jurors?) that he is too dim-witted to have masterminded a shell game of international wire transfers, and that he had no idea what was happening in the closely-held company of which he was Chief Executive Officer. (For an entertaining take on what We The People think of SBF’s disclaimers, see responses in this thread ttps://twitter.com/SBF_FTX/status/1591989554881658880, especially the video posted by “Not Jim Cramer”). 

The word on the street is that his former partner Caroline Ellison (who he has been implicitly throwing under the bus with his disclaimers of responsibility for the multi-billion dollar transfers from his FTX to her Alameda company) may well be cutting a deal with prosecutors to testify against SBF.  It remains to be seen whether SBF’s monumental political donations will suffice to keep him from doing hard time.

But all that legal drama aside, the SBF saga brings up some interesting issues on risk management. Earlier here on EWED James Bailey  highlighted a revealing exchange between SBF and Tyler Cowen, in which SBF displayed a heedless neglect of the risk of catastrophic outcomes, as long as there is a reasonable chance of great gain:

TC: Ok, but let’s say there’s a game: 51% you double the Earth out somewhere else, 49% it all disappears. And would you keep on playing that game, double or nothing?

SBF: Yeah…take the pure hypothetical… yeah.

TC: So then you keep on playing the game. What’s the chance we’re left with anything? Don’t I just St. Petersburg Paradox you into non-existence?

SBF: No, not necessarily – maybe [we’re] St. Petersburg-paradoxed into an enormously valuable existence. That’s the other option.

Boiled down, the St Petersburg Paradox involves a scenario where you have a 50% chance of winning $2.00, a 25% (1/4) chance of winning $4.00, a 1/8 chance of winning $8.00, and so on without limit. If you add up all the probabilities multiplied by the amount won for each probability, the Expected Value for this scenario is infinite. Therefore it seems like it would be rational, if you were offered a chance to play this game, to stake 100% of your net worth in one shot. However, almost nobody would actually do that; most folks might spend something like $20 or maybe 0.1% of their net worth for a shot at this, since the likely prospect of losing a large amount does not psychologically compensate for the smaller chance of gaining a much, much larger amount. But SBF is not “most folks”.

Victor Haghani recently authored an article on risk management and on SBF’s approach:

Most people derive less and less incremental satisfaction from progressive increases in wealth – or, as economists like to say: most people exhibit diminishing marginal utility of wealth. This naturally leads to risk aversion because a loss hurts more than the equivalent gain feels good. The classic Theory of Choice Under Uncertainty recommends making decisions that maximize Expected Utility, which is the probability-weighted average of all possible utility outcomes.

SBF explained on multiple occasions that his level of risk-aversion was so low that he didn’t need to think about maximizing Expected Utility, but could instead just make his decisions based on maximizing the Expected Value of his wealth directly. So what does this mean in practice? Let’s say you find an investment which has a 1% chance of a 10,000x payoff, but a 99% chance of winding up worth zero. It has a very high expected return, but it’s also very risky. How much of your total wealth would you want to invest in it?

There’s no right or wrong answer; it’s down to your own personal preferences. However, we think most affluent people would invest somewhere between 0.1% and 1% of their wealth in this investment, based on observing other risky choices such people make and surveys we’ve conducted…

SBF on the other hand, making his decision strictly according to his stated preferences, would choose to invest 100% of his wealth in this investment, because it maximizes the Expected Value of his wealth.

Even in a game with a fair 50/50 outcome, a player with finite resources will eventually go broke. This is the “Gambler’s Ruin” concept in statistics. SBF’s outsized penchant for risk took his net worth to something like $30 billion earlier this year, something we more-timid souls will never achieve, but it eventually proved to be his undoing.

Most people have a more or less logarithmic sense of the utility of money – if you only have $1000, the gain or loss of $100 is significant, whereas $100 is lost in the noise for someone whose net worth is over a million dollars. SBF apparently felt that he was playing with such big numbers, that he did not need to worry about big losses, as long as there was a chance at a big, big win. Here is a Twitter Thread  by SBF, from  Dec 10, 2020:

SBF: …What about a wackier bet? How about you only win 10% of the time, but if you do you get paid out 10,000x your bet size?

[So, if you have $100k,] Kelly* suggests you only bet $10k: you’ll almost certainly lose. And if you kept doing this much more than $10k at a time, you’d probably blow out.

…this bet is great Expected Value; you win [more precisely, your Expected Value is] 1,000x your bet size.

…In many cases I think $10k is a reasonable bet. But I, personally, would do more. I’d probably do more like $50k.

Why? Because ultimately my utility function isn’t really logarithmic. It’s closer to linear.

…Kelly tells you that when the backdrop is trillions of dollars, there’s essentially no risk aversion on the scale of thousands or millions.

Put another way: if you’re maximizing EV(log(W+$1,000,000,000,000)) and W is much less than a trillion, this is very similar to just maximizing EV(W).

Does this mean you should be willing to accept a significant chance of failing to do much good sometimes?

Yes, it does. And that’s ok. If it was the right play in EV, sometimes you win and sometimes you lose.

(*The Kelly criterion is a formula that determines the optimal theoretical size for a bet.)

Haghani concludes, “It seems like SBF was essentially telling anyone who was listening that he’d either wind up with all the money in the world, which he’d then redistribute according to his Effective Altruist principles – or, much more likely, he’d die trying.”

( Full disclosure: I have lost an irritating amount of money thanks to SBF’s shenanigans. My BlockFi crypto account is frozen due to fallout from the FTX collapse, with no word on if/when I might see my funds again. )

Highlights from EA Global DC

James Bailey3 Comments

I was in DC last weekend for the Effective Altruism Global conference. I met a lot of smart people who are going to have a huge impact on the world, and some who already are. I’ll share a few of my favorite highlights here, with the disclaimer that most quotes won’t be exact:

The mistake every do-gooder makes is coming to a country and thinking ‘I’m just here to help people, I’m not a political actor.’ Guess what? You are. What you do changes the balance of power, often toward the center

Chris Blattman

I’m funding the Yale spit test? The world doesn’t make sense [Yale, NIH, et c should be on it]… its like, if I won an academy award or NBA MVP, how screwed up would the world be?

Tyler Cowen, referring to Fast Grants

You should all be political independents, both parties are terrible. You should be voluntary social conservatives, behave like Mormons…. we need a marginal revolution toward the better parts of the Mormon / social conservative package

Tyler Cowen
“Keep right” indeed

Tyler later specified that the main things he meant by this were to marry young and not drink, though I don’t think he realized how common the latter already is:

As he often does, Tyler recommend that people travel more:

If I meet someone who’s been to 40 countries I tell them they should travel more, and to weirder places

Tyler Cowen

But when someone asked “How much travel is too much”, he came up with this limiting principle:

How much travel is too much travel? 10% after your significant other gets mad at you

Tyler Cowen

I asked Matt Yglesias how much of his policy influence comes just from writing things online, and how much from personal connections and being in DC. He said something like:

Personal connections matter a lot given how real people change their minds, but there’s also less of a dichotomy than you’d think. For instance, a WaPo column of mine was getting passed around the White House, but I wrote it because someone in the WH suggested the topic. Politicians often communicate with each other via the media, though I wish they wouldn’t. Just talk to each other, you work in the same building!

Matt Yglesias

His take on the changing media environment:

My tweets are more influential than my columns & substack, because they are read so much more & I’m followed by many journalists. Overall though now is a great time for specialists, obsessives and weirdos. Construction Physics is a great blog now but if he’d written it in 2003 people would just be like, WTF. On the other hand my [generalist] college blog did well in 2003 but if a college student wrote the same kind of things today people would say, who cares?

Matt Yglesias

Journalists are suspicious haters, that’s our function in society

Can’t remember if this was Matt Yglesias or Kelsey Piper

Tyler and Matt were both telling people that you can accomplish your goals more effectively by being more “normie” in some ways. This can be a bit of a sacrifice, but:

If you can give a kidney, you can learn to tie a tie, give a firm handshake, and look people in the eye

Matt Yglesias

I’m some combination of smart enough and arrogant enough that its normally rare for me to meet someone and think “oh, you’re smarter than I am”. But at EAG it was common; not just because of the ridiculous numbers of top-university degrees and real-world accomplishments, but the breadth and depth of the conversations, everything from mental math to number theory, AI to finance, to a surprisingly convincing pitch for the relevance of metaphysics for political theory.

The other main place I think this is SSC / ACX meetups

It wasn’t a step up for everyone though; I talked to someone at a top hedge fund who said the people he worked with were “are the smartest, most dedicated people I’ve been around…. smarter than EAs, more able to execute than mathematicians at [top PhD program he was at]”. They work 12 hour days, actually working the whole time (no long lunch break, small talk with colleagues, reading social media on their computers)… but all in a ruthless, selfish, impressively successful quest to outsmart the market and make more money.

Overall it was a great time and helped me narrow down my plans for what to do with my time and brainpower post-tenure. If you’re interested there are more conferences ahead.

Boutique Science

James Bailey1 Comment

Science keeps getting bigger- more researchers, more funding, and of course more publications. Scientific progress is much harder to measure, but there are good arguments that it’s roughly flat over time. This implies that productivity per researcher is plummeting.

Source

There’s been a lively debate about what drives this falling productivity- is it that the easy discoveries got made first, leaving only harder ones for today’s scientists? Or is something else tanking scientific productivity, like the bureaucratic way we organize scientific research today?

A recent paper, “Slowed canonical progress in large fields of science“, suggests that the growth in the number of researchers and publications could itself be part of the problem. Comparing scientific fields over time, they find that:

When the number of papers published per year in a scientific field grows large, citations flow disproportionately to already well-cited papers; the list of most-cited papers ossifies; new papers are unlikely to ever become highly cited, and when they do, it is not through a gradual, cumulative process of attention gathering; and newly published papers become unlikely to disrupt existing work. These findings suggest that the progress of large scientific fields may be slowed, trapped in existing canon.

What is driving this? They argue:

First, when many papers are published within a short period of time, scholars are forced to resort to heuristics to make continued sense of the field. Rather than encountering and considering intriguing new ideas each on their own merits, cognitively overloaded reviewers and readers process new work only in relationship to existing exemplars. A novel idea that does not fit within extant schemas will be less likely to be published, read, or cited. Faced with this dynamic, authors are pushed to frame their work firmly in relationship to well-known papers, which serve as “intellectual badges” identifying how the new work is to be understood, and discouraged from working on too-novel ideas that cannot be easily related to existing canon. The probabilities of a breakthrough novel idea being produced, published, and widely read all decline, and indeed, the publication of each new paper adds disproportionately to the citations for the already most-cited papers.

Second, if the arrival rate of new ideas is too fast, competition among new ideas may prevent any of the new ideas from becoming known and accepted field wide.

Supposing they are correct, it’s not totally clear what to do. At the biggest level we could fund fewer researchers in large fields, or push more fields to be like economics, where the quality of each researcher’s publications is valued much more than the quantity. But what can an individual researcher do differently? One idea is “boutique science” or “hipster science”, trying to find the smallest or newest field you could reasonably attach yourself to.

Another idea is that the role of generalists and synthesizers is becoming more valuable, as Tyler Cowen often says and David Esptein applies to science in his book Range. When papers are coming out faster than anyone can read, we need more people to sift through them and explain which few are actually important and which are forgettable or wrong. There are lots of ways to do this- review articles, meta-analysis, replication at scale, and of course blogs. But the junk pile is going to keep growing, so we’ll need new and better ways of finding the hidden gems.

Gifts for a Time of Inflation and Supply Bottlenecks

James Bailey1 Comment

I’ve never been great at gifts, and don’t have much in the way of specific ideas now. But I’ve been thinking about what the macroeconomic environment means for gift-giving.

First, as you’ve probably heard by now from us or elsewhere, if you want to get any physical gift I’d order it now, since shipping is a mess and prices are only going up. I’d especially recommend this for complex electronics that could become hard to find- its part of why I got my wife an iPad for her birthday this summer. Foreign food and drink that can be stockpiled is always a good idea, but perhaps especially now; think wine, Scotch, or Beirao liqueur (a Portuguese drink that was my favorite discovery this year). Wine and liquor make good stores of value in an time of inflation.

Alternatively, you could avoid the scarcity of physical goods by turning to the digital realm. If your economistic heart yearns to give cash, consider giving some your favorite stock or cryptocurrency instead- its both more personalized and less subject to inflation. Or if you think you can judge the recipient’s taste well enough, subscribe them to one of your favorite Substacks or podcasts. My recommendations:

Unsupervised Learning, Razib Khan– Genomics and History

The Diff, Byrne Hobart– Finance and Strategy

Astral Codex Ten, Scott Alexander– Rationality, Psychiatry, Everything

The Readers Karamazov– Funny podcast on literature and philosophy (now seems entirely free though)

Or if you really have money to burn, go for the Bloomberg subscription. I always run out of free reads on the Tyler Cowen articles and so can’t read Matt Levine, even though he has the magic ability to teach you finance while making you laugh. But the subscription is expensive and Mike Bloomberg doesn’t need the money, while the Substacks are relatively cheap and enable talented writers to spend a lot more time writing instead of needing to focus on a real job.

Tyler Cowen, Talent Curator

James Bailey1 Comment

Everyone else at EWED has been too classy (or earnest?) to post it, since it would implicitly be bragging.

But I’m home with a quarantined kid today and need the win. So here is biotech founder Tony Kulesa‘s article on how Tyler Cowen is the Best Curator of Talent in the World.

Highlights:

Tyler has identified talent either earlier than or missed by top undergraduate programs, the best biotech startups, and the best biotech investors, all without any insider knowledge of biotech. In comparison, Forbes 30U30, MIT Tech Review TR35, or Stat Wunderkind, and other industry awards that highlight talent are lagging indicators of success. It’s hard to find an awardee of these programs that was not already widely recognized for their achievements among insiders in their field. The winners of Emergent Ventures are truly emergent. 

What explains Tyler’s ability to do this?

1. Distribution: Tyler promotes the opportunity in such a way that the talent level of the application pool is extraordinarily high and the people who apply are uniquely earnest. 

2. Application: Emergent Ventures’ application is laser focused on the quality of the applicant’s ideas, and boils out the noise of credentials, references, and test scores. 

3. Selection: Tyler has relentlessly trained his taste for decades, the way a world class athlete trains for the olympics. 

4. Inspiration: Tyler personally encourages winners to be bolder, creating an ambition flywheel as they in turn inspire future applicants.

This seems right as far as it goes, and there is more depth in the article, but there has to be more to the story than we can see from the outside. Luckily Tyler has said he is writing a book on identifying talent.