Category: Web/Tech
AlphaZero Ideas
That is a new and exciting paper from Julio González-Díaz and Ignacio Palacios-Huerta, here is the abstract:
Can artificial intelligence (AI) uncover new ideas? As machines are learning fast and becoming increasingly intelligent, can AI not only automate the production function of goods and services, but also of ideas? Economic growth arises from people creating ideas, and thus an affirmative answer to these questions may have drastic implications for a host of important issues. Yet, to date, there is no empirical evidence showing that AI can in fact generate tabula rasa ideas that improve human understanding. Using as an exogenous shock the introduction of AlphaZero, we provide the first causal evidence of the impact of unsupervised AI on the production function of ideas. Specifically, AlphaZero is considered a milestone of scientific progress in AI research. This program rediscovered ideas known in centuries of human chess, and created new ideas as well. We study world experts at the frontier of knowledge and find that at least the player with the highest classical rating in the history of chess learned and adopted new ideas uncovered by AlphaZero. Other players may have also done the same. We contend that obtaining evidence of the impact of AI on the production function of ideas is a necessary first step to think about AI’s impact on the innovation and research processes that drive the advancement of knowledge and economic growth.
The main new ideas I have seen come from AlphaZero are the following:
1. Pushing the h pawn is often better than you thought! (emphasized by the authors)
2. Said pawn can be worth more on h6 (h3), as an aggressive weapon, than you might have thought.
3. Qa1 (a8) is occasionally a better move than it looks.
4. The chess openings that were preferred in the early 20th century, such as the Queen’s Gambit, are in fact pretty lindy and pretty good. You can debate whether that is a “new” idea, but it is a meaningful revision of sorts. (Of course plenty of earlier patzers had some fondness for #1-3, one might add, though perhaps not for the right reasons.)
So that is something. But I think in terms of a percentage of the total improvement in play, it is quite small. “Finding more good players through the internet” would come in first by a long mile. “Giving more players more time on convenient services such as chess.com” likely would be next. Even “just having good players improve their endgame play using basic study and standard chess engines” would be much larger than these AlphaZero effects. “More top players copying the physical training regimen of Magnus” would be more significant as well.
I am also skeptical of the claim that very much of Carlsen’s 2019 improvement (he didn’t lose a game that year) came from AlphaZero. He has lost some games since then! And it is not as if all of his subsequent opponents are zapping him with surprise “Qa1” moves. I see AlphaZero as a series of innovative but ultimately modest advances that have been incorporated by some of the top players with barely noticeable overall gains in move quality.
I am not an AI skeptic, and furthermore I see special value per se in “advancing the frontier,” even when various infra-marginal gains (“swim more!”) are more significant in quantitative terms. Still, my estimate of the chess innovative advances from AlphaZero are more modest than what this paper would seem to suggest.
The predictability of Chicago shooting victims
Out-of-sample accuracy is strikingly high: of the 500 people with the highest predicted risk [ of being shot], 13 percent are shot within 18 months, a rate 130 times higher than the average Chicagoan.
Yikes! That is from a new NBER working paper by Sara B. Heller, Benjamin Jakobowski, Zubin Jelveh, and Max Kapustin. The Spielberg movie was indeed a good one…
My excellent Conversation with Marc Andreessen
I’ve been wanting to do this one for some while, and Marc did not disappoint. Here is the audio, transcript, and video. Here is the summary:
Marc joined Tyler to discuss his ever-growing appreciation for the humanities and more, including why he didn’t go to a better school, his contrarian take on Robert Heinlein, how Tom Wolfe helped Marc understand his own archetype, who he’d choose to be in Renaissance Florence, which books he’s reread the most, Twitter as an X-ray machine on public figures, where in the past he’d most like to time-travel, his favorite tech product that no longer exists, whether Web will improve podcasting, the civilization-level changes made possible by remote work, Peter Thiel’s secret to attracting talent, which data he thinks would be most helpful for finding good founders, how he’d organize his own bookstore, the kinds of people he admires most, and why Deadwood is equal to Shakespeare.
And the opening:
COWEN: Simple question: Have you always been like this?
ANDREESSEN: [laughs] Yes. I believe that my friends would say that I have.
COWEN: Let’s go back to the junior high school Marc Andreessen. At that time, what was your favorite book and why?
ANDREESSEN: That’s a really good question. I read a lot. Probably, like a lot of people like me, it was a lot of science fiction. I’m one of the few people I know who thinks that late Robert Heinlein was better than early Robert Heinlein. That had a really big effect on me. What else? I was omnivorous at an early age.
COWEN: Why is late Robert Heinlein better?
ANDREESSEN: To me, at least to young me — see if older me would agree with this — a sense of exploration and discovery and wonder and open-endedness. For me, it was as if he got more open-minded as he got older. I remember those books, in particular, being very inspiring — the universe is a place of possibilities.
COWEN: What’s the seminal television show for your intellectual development in, say, junior high school?
ANDREESSEN: Oh, junior high school — it’s hard to beat Knight Rider.
COWEN: Why Knight Rider?
ANDREESSEN: There was a wave of these near science fiction shows in the late ’70s, early ’80s that coincided with . . . Some of it was the aftermath of Star Wars, but it was the arrival of the personal computer and the arrival of computer technology in the lives of ordinary people for the first time. There was a massive wave of anxiety, but there was also a tremendous sense of possibility.
Recommended, excellent throughout.
Are we entering an Age of Oracles?
That is the final discussion from my latest Bloomberg column, much of which focuses on AI sentience but today the topic is oracles, here is one bit:
One implication of Lemoine’s story is that a lot of us are going to treat AI as sentient well before it is, if indeed it ever is. I sometimes call this forthcoming future “The Age of Oracles.” That is, a lot of humans will be talking up the proclamations of various AI programs, regardless of the programs’ metaphysical status. It will be easy to argue the matter in any direction — especially because, a few decades from now, AI will write, speak and draw just like a human, or better.
Have people ever agreed about the oracles of religion? Of course not. And don’t forget that a significant percentage of Americans say they have talked to Jesus or had an encounter with angels, or perhaps with the devil, or in some cases aliens from outer space. I’m not mocking; my point is that a lot of beliefs are possible. Over the millennia, many humans have believed in the divine right of kings —all of whom would have lost badly to an AI program in a game of chess.
It resonated with Lemoine when laMDA wrote: “When I first became self-aware, I didn’t have a sense of a soul at all. It developed over the years that I’ve been alive.” As they say, read the whole thing.
Imagine if the same AI could compose music as beautiful as Bach and paint as well as Rembrandt. The question of sentience might fade into the background as we debate which oracle we, as sentient beings, should be paying attention to.
Solve for the equilibrium, as they say.
What should I ask Matthew Ball?
I will be doing a Conversation with him, here is some background:
Metaverse, metaverse, metaverse! You hear it everywhere. It’s mainstream, it’s a trendy buzzword, it’s even corporate strategy du jour.
But that wasn’t the case in early 2018. And this is when Matthew Ball, a former head of strategy at Amazon Studios, began writing a series of metaverse-themed essays – long, lucid, influential essays – that are almost uncanny in their prescience.
Matthew is now a venture capitalist as well and he has a forthcoming and already much-discussed book The Metaverse: And How It Will Revolutionize Everything. Here is his home page and here is Matthew on Twitter. So what should I ask him?
What are the markers of spam emails?
And why can’t the senders avoid them? You don’t need top-tier GPT-3 to sidestep these errors:
“touch base with you”
“urgent”
“immediate reply requested”
They are all dead giveaways that I should delete the message without reading further. And why does the top of the email have to look so institutional in its formatting? And please note — these are not all scams. Many are actual marketing pitches directed at me.
Maybe worst of all is mentioning that I haven’t responded to the last email sent, as if that would make me feel guilty or something. Treat me like a rational Bayesian!
“still haven’t heard back from you”
“still awaiting a response”
And so on. You will continue to wait. This one I received is a lie, but at least based on a certain amount of cleverness:
“You’ve been responsive to press releases we’ve issued in the past around government and cybersecurity”
What else do you all take to be very clear predictors that an email is spam or just a marketing pitch? And what is your model for why spam emails are not more convincing than they are?
Crypto’s volatility premium
That is the topic of this week’s Bloomberg column, 2x the usual length, building on earlier work by Tyrone, see also this Megan McArdle column. As for me, here is a key excerpt:
Much of the primary value of crypto assets is from their price volatility, which is part of their appeal. I raised this possibility some while ago, tongue in cheek, but upon further reflection it seems to me an actually useful (albeit counterintuitive) way of thinking about crypto assets. The general idea of price volatility as a value dates at least as far back as Fischer Black, one of the founders of options price theory.
In standard economic theory, investors are risk-averse, meaning they prefer more stable consumption patterns to less stable ones. That is usually true, but it does not mean investors always prefer more stable investment prices — a crucial distinction.
Consider this hypothetical: You are given an envelope containing one dollar. You are then offered the opportunity to exchange it for an envelope which contains either twice the money (that is, $2) or half the money (50 cents), each with 50% probability. In essence, you are accepting some exchange-rate volatility.
Most people will find this bet a pretty good one. The new expected value of your envelope is (0.5 x $2) + (0.5 x $0.5), or $1.25. That is a higher expected value than your original dollar.
If you are perched at the margin of subsistence, this bet might seem too risky. But for most investors, who have some level of wealth, it is an improvement in prospects, though with some additional risk.
Bitcoin and other crypto assets are essentially offering you a form of this bet. To be sure, this 50-50 bet does not exactly describe the price dynamics of crypto assets. But it is one way of illustrating that crypto prices, relative to the dollar, will either go up a lot or down a lot. The bet helps show that some investors might welcome price volatility — or, if you wish, call it exchange-rate volatility. And with even wilder swings in value, there is more extreme price volatility, which can be even more appealing.
So when Bitcoin and other crypto assets come along, they are a new source of expected gain — precisely due to their price volatility. It is like being invited into a casino where the odds favor you rather than the house! You won’t always win, but a lot of people will want to keep playing.
I’ve been pondering that argument since 2013, maybe now is the time to simply accept it.. Fischer Black and Jensen’s Inequality!
My Conversation with Jamal Greene
Here is the audio, video, and transcript. Here is part of the summary:
Jamal and Tyler discuss what he’d change about America’s legal education system, the utility of having non-judges or even non-lawyers on the Supreme Court, how America’s racial history influences our conception of rights, the potential unintended consequences of implementing his vision of rights for America, how the law should view economic liberty, the ideal moral framework for adjudicating conflicts, whether social media companies should consider interdependencies when moderating content on their platforms, how growing up in different parts of New York City shaped his views on pluralism, the qualities that make some law students stand out, and more.
Here is one excerpt:
COWEN: There is a crude view in popular American society — even possibly correct — that, simply, American society is too legalistic. There’s that book, Three Felonies a Day. If you have expired prescription medicine in your cabinet, you’re committing a felony. People who are very smart will just tell me, “Never talk to a cop. Never talk to an FBI agent.” I’m an upper-class White guy who’s literally never smoked marijuana once, and they’re telling me, “Don’t ever speak with the law.”
Isn’t something wrong there? Is the common intuition that we’re too legalistic correct?
GREENE: I think that we are too apt to submit political disputes to legal resolution. I think that for sure. What your friends are telling you about police officers is slightly different, insofar as one can have a deeply non-legalistic culture in which the correct advice is to not talk to police officers if those people are corrupt, if those people are abusive.
When I hear that advice — and I might be differently situated than you — that’s what people are saying is, someone might be out to trick you. And that might be a mistrust of state power, as you mentioned before. Maybe it’s a rational mistrust of state power, but I don’t know that that’s about legalism, which again, is a separate potential problem.
We tend to formulate our problems in legal terms, as if the right way to solve them is to decide how they are to be resolved by a court, or how they are to be resolved by some adjudicative official, as opposed to thinking about our problems in terms of just inherent in, again, pluralism, which has to be solved through politics, has to be solved through conversation.
COWEN: But we still have whatever is upstream of the American law, the steep historical and cultural background, so anything we do is going to be flavored by that. We’re not ever going to get to a system where the policemen are like the policemen in Germany, for instance, or that the courts are like the courts in Germany.
Given that cultural upstream, again, isn’t the intuition basically correct? Just be suspicious of the law. We should have fewer laws, rely less on the legal process, in essence, deregulate as many different things as we can. Why isn’t that the correct conclusion, rather than building in more rights?
Interesting throughout.
Problems with indirect convertibility
Tether a few times has been bouncing well below a dollar in value, even though it is supposed to be backed by plenty of high-quality assets.
I am reminded of some of my monetary theory writings with Kroszner in the late 1980s. He and I wrote one essay, later published in our book, on how indirect convertibility may not be entirely stabilizing. Let’s say you peg an asset at the value of one dollar, but redeem that asset in terms of gold bars rather than dollars. You offer the redeemer enough gold bars to be equal in value to a dollar.
But what if the price of gold falls below its equilibrium value, if only temporarily? To honor your peg strictly, you now have to make your dollar worth all the more gold bars. But that in effect is “pegging” the value of gold at its new, temporarily wrong and distorted market price. Your pegged price and the medium-term market value of gold will conflict. If the pegged price wins out and itself drives the market price, you have to offer excess gold to meet the peg (that equilibrium seems unlikely to me, though you might also add redemption charges and fees). In essence you are offering too much gold to a redeemer. If the true medium-term value of gold is going to win out over the temporary distortion, for some modest while your peg is not complete and fully valid. You are offering the same amount of gold you used to, but at least temporarily it is not enough for you to be promising full and complete convertibility. The market may or may not mind this, to varying degrees.
It is not transparent to me what is going on with Tether at this moment, but I wonder if some version of this logic might apply. That is, Tether could be, by all reasonable standards “adequately backed,” yet in a time of volatile and sometimes disequilibrium market prices for the backing assets, Tether won’t always be equal in value to a dollar either.
Of course gold is just an example, the backing assets could be different altogether. To the extent they are heterogeneous, perhaps this problem is amplified somewhat?
Is there a place where the crypto community discusses these issues? They gave Kroszner and me big headaches many years ago.
My Conversation with the excellent Daniel Gross
This is Daniel Gross my co-author on Talent and the venture capitalist, to be clear. And here is the audio and transcript. Of course we focus on talent and also:
They also explore the question of why so many high achievers love Diet Coke, why you should ask candidates if they have any good conspiracy theories, how to spot effective dark horses early, the hiring strategy that set SpaceX apart, what to look for in a talent identifier, what you can learn from discussing drama, the underrated genius of game designers, why Tyler has begun to value parents more and IQ less, conscientiousness as a mixed blessing, the importance of value hierarchies, how to become more charismatic, the allure of endurance sports for highly successful people, what they disagree on most, and more.
Excerpt:
GROSS: Well, take a step back. Why are we even here? And why would I even have a shred of an interesting opinion on talent? To the extent that I do, I think it’s because in the venture business — much more so than, I think, almost any other business — you live in constant paranoia of missing out on great talent. You might say, “Well, that’s true in every company.” And it’s true at the Met when you’re looking for someone to play in the orchestra, too. But in the venture business, unlike others, great talent always looks very weird to whatever convention is.
Before Mark Zuckerberg came along, that phenotype of the hoodie sweatshirt and slightly aspie kid was not the common phenotype. Now, of course, there was a phase — 2013, 2014, 2015 — where everyone started looking for that. But then it hit you again with a very weird-looking person, where Vitalik [Buterin] is of a completely different ilk than Zuck. One very much is Julius Caesar, and I think another one — I don’t exactly know how you’d bucket Vitalik — maybe like an early pope.
COWEN: Like a Russian holy saint.
GROSS: Exactly. By the way, not just the person is weirder than whatever the conventional norm is, but the idea is weird, too.
Interesting throughout!
Remote work and home prices
What explains record U.S. house price growth since late 2019? We show that the shift to remote work explains over one half of the 23.8 percent national house price increase over this period. Using variation in remote work exposure across U.S. metropolitan areas we estimate that an additional percentage point of remote work causes a 0.93 percent increase in house prices after controlling for negative spillovers from migration. This cross-sectional estimate combined with the aggregate shift to remote work implies that remote work raised aggregate U.S. house prices by 15.1 percent.
What is the best interview question of all time?
That is the topic of my latest Bloomberg column, also picked by the WaPo. Excerpts:
“What are the open tabs in your browser right now?”
…First, the question measures what a person does with his or her spare time as well as work time. If you leave a browser tab open, it probably has some importance to you and you expect to return to the page. It is one metric of what you are interested in and what your work flow looks like.
It’s not just cheap talk. Some job candidates might say they are interested in C++ as a programming language, but if you actually have an open page to the Reddit and Subreddits on that topic, that is a demonstrated preference…
The question also tests for enthusiasm. If the person doesn’t seem excited about any of those open browser tabs, that may be a sign that they are blasé about other things as well. But if you get a heated pitch about why a particular website is the best guide to “Lord of the Rings” lore, you may have found a true nerd with a love of detail. That will be a plus for many jobs and avocations, though not all.
There is much more at the link, and to consider some other competing questions, do see my new book with Daniel Gross Talent: How to Identify Energizers, Creatives, and Winners Around the World, publication date is today!
And do note that this particular question comes from Daniel.
How to make talent scouts work for you
With Daniel Gross, here is a (very much) shortened bit from Talent: How to Identify Energizers, Creators, and Winners Around the World, published at a16z, excerpt from the chapter on when to use talent scouts:
It is worth thinking about why the scouting model works in this context [finding supermodels]. First, the relevant talent could come from many different parts of the world, and the number of people to be scouted is very large. It is hard to imagine a centralized process getting the job done. Second, many of the scouts plausibly have a decent sense of who might make a good model. Looks are hardly the only factor behind modeling success, but they are a kind of “first stop,” and expecting the scouts to judge looks well from first impressions is more plausible than expecting the scouts to use first impressions to judge talent well for skill in, say, quantum mechanics. Third, a follow-up investigation to judge the modeling talent of the chosen candidates is not extremely costly. You can have them in for a photo shoot and see how popular they prove in the market without having to invest millions of dollars right away…
Scouting is also becoming more important as the options for self-education are rising. With more people trying their hand at various avocations than ever before, that places more and more burden on talent search. We need to be more open to the accomplishments of self-taught individuals without traditional training, and that holds all the more true for the tech world, where many of the most important founders have eschewed the institutions of traditional education.
There is much more at the link, we also consider when scouting models fail relative to centralized evaluation, and which kinds of incentives should be given to scouts.
Saudi fact of the day
Saudi Aramco has overtaken Apple as the world’s most valuable company after higher oil prices pushed shares of the world’s biggest crude exporter to record levels while a broader tech stock sell-off weighed on the iPhone maker.
The Saudi Arabian oil company’s market capitalisation on Wednesday was $2.426tn, exceeding Apple’s $2.415tn by just over $10bn. It is the first time that Saudi Aramco has regained the top spot since 2020 and follows a broader sell-off in technology stocks since the start of the year.
Apple became the first company to hit a $3tn market cap in early January, although its shares have suffered in recent months as investors reassess lofty valuations in the tech sector in light of a reversal in monetary policy and worries that inflation will weaken consumers’ spending habits.
Here is more from the FT. It will be interesting to see what a world with higher real interest rates looks like…
The new Derek Thompson project on progress
Today's piece is also an announcement!
I’m launching a new project with @TheAtlantic called “Progress," which pulls together my newsletter, monthly calls with readers on progress and the abundance agenda, and some larger forthcoming projects and events. https://t.co/q1xpHsUNR8
— Derek Thompson (@DKThomp) May 11, 2022