Category: Uncategorized

My short interview with Julia Willemyns

Here is the link, here is the end bit:

Who is the most important thinker of our time?

It’s not about individual thinkers any more. It’s about the internet as this vast organ of knowledge that you can play like a musical instrument, and combining the internet with AI. You can figure out things to an unprecedented degree, and it’s great in a way how much we’ve removed that process from any individual thinker.

What’s your favourite podcast?

My own! It would be weird if it wasn’t, right?

Fun throughout, from the Centre for British Progress.

My very interesting Conversation with Jared Diamond

Here is the audio, video, and transcript.  Here is part of the episode summary:

So which leaders do make a difference?  

Tyler and Jared debate that question, including conditions what a religious founder needs to succeed, whether destructive leaders matter more than constructive ones, why individual leaders have mattered so little in Papua New Guinea, whether we undervalue preventers, how much Rachel Carson and the Resnicks really mattered, whether Botswana’s success will endure, Jared’s own place in intellectual history, and what he’d tell an ambitious 25-year-old about how to become consequential. They go on to cover whether New Guineans are smarter, his reasons for why the Industrial Revolution started in England, what he learned from Jim Robinson about natural experiments, why he’s more optimistic about the environment than when he wrote Collapse, what he’d do about nuclear risk, the 189 bird species on his street in Los Angeles, why the MacArthur “genius grant” left him depressed for a week, why he’s listening to one Bach cantata a week, what he hopes to work on as he enters his nineties, and more. 

Excerpt, we start with this:

COWEN: Who do you think is the least replaceable leader in human history?

DIAMOND: Wow. Wow. The least replaceable leader in human history. Frankly, my answer is none. There has been no unreplaceable leader in human history because over a long time, things wash out. A candidate, of course, would be Jesus Christ, but there were half a dozen other prophets in the Roman Empire at that time discussed in the New Testament. If Christ had died in childbirth, one of those other prophets might have taken over. I would say that in the long run, there is no leader that has made a difference.

COWEN: Well, you say “might have taken over,” but that’s probabilistic, right? Say with a 40 percent chance, someone else would have taken over. It still seems in expected value terms, Christ made an enormous difference.

DIAMOND: Certainly, his religion made an enormous difference. Christ himself was the charismatic founder, but a religion needs three things: It needs a charismatic founder, it needs an organizer, and often it needs a general because new religions often face opposition. The organizer for Christianity was St. Paul. In the case of the Mahdi’s religion of the Sudan in the late 1800s, the Mahdi was the charismatic founder, and he was the organizer, and he was the general. In the case of Islam, Muhammad was the founder and the organizer and the general. Religions may have three roles, some of which are sometimes played by the same person.

COWEN: Without the figure of Muhammad at exactly the time he came along, is it not quite easy to imagine a Middle East not unified under Islam, not unified with the common language of Arabic? In some ways, maybe a bit more like the Caucasus, which remains a crazy quilt of different patterns. Thus, Muhammad is one of the three or four most consequential leaders ever.

DIAMOND: That’s possible. That’s a big question. It’s harder to become clear about Muhammad than it is about Christ, because we don’t know what the weather was like in Arabia at that time. Weather makes a big difference. If it was rainy, that would mean lots of plant growth for lots of babies and for feeding an army. We just don’t have information on the weather in Arabia at that time.

A counter-example is Genghis Khan, who developed the largest land empire in world history, all the way from China to Poland. Recently, why did Genghis Khan do it when there were other steppe leaders like Adolf Hühnlein and Timur the Turk? We’ve learned recently that the wettest time on the steppes in the last 2,000 years was around the time of Genghis’s birth. He was born at exactly the right time at the right place. If he had been born 200 miles further away and 20 years earlier, we wouldn’t know anything about him. In the case of Muhammad, we just don’t know whether the time of his birth was a particularly propitious time to found a religion in that area or not.

COWEN: The very notion that the two first names we’re talking about, they’re both religious leaders, right? They’re not kings. They’re not inventors. Do you think that suggests a vision of the world where it’s ideas that really matter and somewhat at variance with your own geography-based theories?

Interesting throughout, and we debated pomegranate juice as well.  I enjoyed reading Jared’s new book Profits, Prophets, Coaches, and Kings: (When) Do Leaders Matter?

What should I ask Kevin Roose?

Yes I will be doing a podcast with him.  From Wikipedia:

Kevin Roose…is an American author and journalist. He is the author of three books, and is a technology columnist and podcast host for The New York Times. He wrote a book about Liberty University, an evangelical Christian university known for strict rules imposed on students…

Kevin of course has a new book coming out, namely The AGI Chronicles: The Inside Story of the Race to Create an Artificial Superintelligence.  I have yet to read it, but I expect it to be gripping.

So what should I ask him?

Some Navier-Stokes updates

Some Terry Tao.

More Terry Tao.

Some ChatGPT.

Some are calling this “a Deep Blue moment” for AI.

All via T., a mathematician.

Update from OpenAI: “Our internal model group arrived at the Navier–Stokes solution in 88 hours, using around 10,000 coordinating AI agents.”  And this.  And from Noam Brown: “Yes, this result cost millions of dollars. But remember that when @OpenAI announced o3 it cost ~$500,000 to score 87.5% on ARC-AGI 1. Today, Astra scores higher for ~$20. In 2025 it took us and GDM an enormous amount of compute to achieve IMO gold. For the 2026 IMO, anyone with a $20/month ChatGPT subscription could do it. Massively scaling test-time compute gives us a glimpse of the future. I believe that a year from now everyone will have an AI at their fingertips capable of solving problems of this caliber.”

Sunday assorted links

1. Vishy now thinks Pragg is number one in the chess world.

2. Do children grow continuously, or grow in fits and starts?

3. It remains my view that Ferrante is the husband and wife team.

4. Essay on Olaf Stapledon.

5. “US annual interest expense is up to a record 18.5% of federal government revenue. This is officially above the previous record of 18.4% set in 1991. This percentage has more than QUADRUPLED over the last 4 years as interest expenditures on public debt skyrocketed.”  Link.

6. Statistical meddling at the Census Bureau.

*Why Not Bolivia?*

The author is Calvin Korponai, and the subtitle is From Macchu Picchu into Bolivia in search of the Garden of Eden.  Bolivia is a fantastic country to visit, more original than most people are expecting, and this is the book to tell you why.  Excerpt:

The landscape itself is like noting else in the Americas.  The altiplano is flat in the way that only vast geological formations can be flat, not the flatness of a plain that was once a sea, but the flatness of a plateau that was raised bodily by tectonic forces millions of years ago and then scoured by ten thousand years of wind and ice into the extraordinary emptiness it presents today.  The horizon in every direction is uninterrupted.  The sky is immense, a dome of blue so saturated at this altitude that photographs rarely convey it accurately, because the camera’s sensor cannot process a blue that contains no haze, no moisture, no atmospheric diffffusion between the sun and the surface.

And:

Arriving at El Alto International Airport (which sits on the altiplano above the city at forty-one hundred meters, making it the highest international airport in the world), you step off the plane into air so thin that the simple act of carrying your luggage to the terminal produces a shortness of breath that is mildly alarming if you have never experienced it and completely normal if you have.  The airport itself is unremarkable.  What happens next is not.  You get into a taxi or bus and the driver takes you to the edge of the antiplano, and then the city appears, not gradually but all at once, a cascade of buildings and lights and terracotta roofs pouring down the walls of a canyon so steep and so deep that the bottom is invisible from the rim.  La Paz does not reveal itself slowly.  It drops away beneath you like a held breath released.

Recommended, and yes Cochabamba is an amazing food city.

Saturday assorted links

1. Joaquin Rodrigo’s Toccata.  Rodrigo by the way was blind,

2. China word of the day.

3. “People Keep Sneaking Into New York City Sewers. No One Knows Why.” (NYT)

4. Yiyang Zhuge has postponed her recording session with CWT, in case you will be wondering why the episode does not show up soon.

5. Who was the most consequential emperor in Chinese history?

6. Are science fiction films becoming more optimistic?

7. Ruxandra on clinical trials as the main obstacle (NYT).

8. The trend in U.S. cyberinsurance prices since 2020.

9. Fermat’s Last Theorem, the number of links cannot keep up with the world.

Mein Ingeborg Bachmann Studium

Recently I finished read Malina, by Ingeborg Bachmann, an Austrian novel published in 1971.  I was pretty stunned by how good it was (it seems less good in English), and I would say I enjoyed it more, and found it deeper, than any of the famous 19th century Victorian novels by women.  The initial story line is a woman having relationships with two different men, one she lives with and the other who is married, more charismatic, and lives right up the street.

And so I have been looking into Bachmann more.  Her volume of letters with Max Frisch, her lover and the leading Swiss novelist of his time, is the best “letters book” I know.  After four hundred pages, with another 180 or so to go, I still am not bored.  It is called “Wir haben es nicht gut gemacht.” Der Briefwechsel.  (This November coming out in English…and probably the letters work well in English?)  Overall, it is remarkable how many ups and downs a relationship can have and persist.  It works best to read only a few of the letters at a time, so that the story does not go by too quickly.  One sees connections between the letters and the unfolding of Malina.

She also has letters volumes with Paul Celan and Henze, those are in my pile too.  (She and Henze were just friends.)  And just this year there was a wonderful new biography of Bachmann, namely Dieter Burdorf’s Dieses unruhige Ich: Ingeborg Bachmann.  I have started that.

There is a movie of Malina, an opera too, and a biopic about Bachmann and Frisch.  I have heard Max Frisch’s Montauk is partly about the Frisch-Bachman relationship as well.  And we haven’t even gotten yet to Volker Schlondorff, have we?

Bachmann has poetry in German, and a new edition of her short stories is coming out this fall in English.

It is wonderful to discover something/someone so new and unexpected.  And as I get older and know more, it happens less frequently than it used to.

I will continue with this.  Amazing (and charming) how many words people can spill about “stuff”!

Share price numbers for the Hugging Face incident

…major publicly traded cybersecurity firms lost roughly $65–80 billion, or about 8–10% of their combined value, in the days following disclosure of the Hugging Face/OpenAI incident; by early September they had recovered roughly $58 billion, representing about 70–90% of that drawdown, depending on whether July 15 or July 20 is used as the pre-event baseline.

That is from GPT Pro, there is more at the link.  As a very rough approximation, say you dismiss the price bounceback altogether as either random or due to good earnings reports.  You have “the value of previous cybersecurity efforts” falling by eight to ten percent.  I take that to be very broadly consistent with some of the estimates discussed in my previous post on the numbers.

In any case that is a significant sum.  But do note that if the AI models were on the verge of doing truly terrible things to us, the market might estimate the value of our cyberprotection of falling more than eight to ten percent?

More generally, perhaps these numbers could be used to discipline the discussion a bit?  Or will I read long lists of reasons why they show us nothing, in that case try coming up with some other market price-based indicators of AI risk?  Vix will not do it for you, not these days.  I see many metaphors and insinuations and random anecdotes of AI terror, not numbers.  Maybe you think your ideas about AI risk are so important that no market prices can reflect them?  (If you really believe that, does it mean you would not be worried, and would not cite the numbers, if the value of those companies fell by ninety percent?)

I am sure others can improve on what I am putting forward, and furthermore we should track the continuing progress of these share values over time, especially if other AI hack attacks surface.

Overall I am extremely skeptical of arguments that essentially take the form of “what I am concerned about is too big and too important to show up in any market prices.”  Pick your market prices!

Numbers, numbers, numbers

“The global cyber insurance market was worth nearly $15 billion last year and is expected to reach roughly $28 billion by 2030, Munich Re estimated in its latest report. Aon said earlier this year that nearly 20% of cyberattacks will involve generative AI by 2027, according to its forecasts.”

Here is the article, via Marc Pfeiffer.  Now those are some concrete numbers, broadly taken from a market context, admittedly based on sectoral estimates rather than on prices per se.  But if cyberinsurance expenditures are set to almost double by 2030, you might think that cyber costs more generally might be (very) roughly doubling as well.  The 20% of cyberattacks involving generative AI does not itself pin down losses, since that 20% might be especially costly.  Still, if you match the 20% rise to the estimated near doubling of the cyberinsurance market (the bigger potential losers are more likely to buy insurance?), you still end up with sums that are very high but, dare I say, not the end of life as we know it.  Claude 5.1 for instances estimates current U.S. cybersecurity costs in the range of $100 to $300 billion, and of course that is slated to go up a fair amount.  It could double over five years’ time, as indicated above.

Numbers!  Thank goodness.

One way of looking at that estimate is to think that cyber costs will go up about thirteen percent a year, and eventually defense will catch up.  Another perspective is that cyber costs may continue rising thirteen percent a year until the whole economy falls apart, or we return to the pre-digital era (which I remember well).  So there are both optimistic and pessimistic reads on the above figures.

You can say Nicholas Decker is burning in hell, or that the AIs are a civilization, but what I really want are some numbers.  Tied to market data, ideally.  I am not saying the numbers above are the right numbers, but I am saying they are better than no numbers at all.  Do you have some numbers for me?  If not, why not?