iiinova — where your digital self keeps evolving

Carl T. Bergstrom @carlbergstrom.com · 20h

I struggled with that. What’s he doing?

110

Carl T. Bergstrom @carlbergstrom.com · 04/10/2026

they don’t know the red gape means this is a juvenile corvus corax

Me, an older white guy holding a glass while wearing a Willie Watson concert T-shirt with a raven perched on the skull of a bighorn sheep

41465

Carl T. Bergstrom @carlbergstrom.com · 03/10/2026

Bizarrely, this is one of the six graphs illustrating "just how much we need A.I.".

Graph of wages and salaries
Graph of after-tax corporate profits

715531

Carl T. Bergstrom @carlbergstrom.com · 03/10/2026

The next five graphs are even less convincing, and the argumentation becomes even weaker. It was almost impossible for me to imagine how a seemingly intelligent person could have written it—but I found a secret clue buried in the text itself.

A.I. has the potential to enhance the way we live in the same way. It saved me and my colleague, Will McGrew, many hours of prowling the internet to research this essay. That left us with time to do other tasks.

1637347

Carl T. Bergstrom @carlbergstrom.com · 03/10/2026

And talk about having it both ways. Solow was wrong about computers not increasing productivity, Rattner argues, but we need AI because productivity didn't increase. Similarly, if Solow was so completely wrong, why didn't the gains of the 90s persists?

In 1987, the economist Robert Solow, who would later win a Nobel Prize, said, “You can see the computer age everywhere but in the productivity statistics.” He spoke prematurely. Productivity rose sharply in the late 1990s, stimulating growth, adding jobs and leading the federal government under President Bill Clinton to an unusual (and never matched) string of budget surpluses.

31377

Carl T. Bergstrom @carlbergstrom.com · 03/10/2026

And why are the 50 years are the anomaly, not the shorter early post-war period? Meanwhile Rattner must know how feeble it sounds to ascribe postwar US growth to delayed benefits of electricity rather than the postwar world order, government investment in R&D, women entering the workforce, etc...

More will come. History suggests that innovation’s effects on the economy can take years, even decades, to play out. The surge in productivity growth after World War II was driven in part by the continuing deployment of electricity generators and other 19th-century inventions.

417513

Carl T. Bergstrom @carlbergstrom.com · 03/10/2026

Yeah. Exactly. Evidence that AI will increase productivity. Instead we get a few cherry-picked anecdotes.

Examples of its power already abound. Siemens uses its A.I. co-pilot to walk technicians through repairs of factory machines, decreasing time spent on unplanned equipment maintenance by 25 percent. Kaiser Permanente has deployed an A.I. scribe to streamline note-taking, cutting documentation time for physicians by nearly 16,000 hours. John Deere has an A.I.-equipped camera system that saved farmers 31 million gallons of herbicide mix in 2025 by spraying the substance only where the system detects weeds.

81764

Carl T. Bergstrom @carlbergstrom.com · 03/10/2026

Take the first chart. The argument here is that productivity increases have been low in the US since the postwar period, so we need AI. See anything missing?

514310

Carl T. Bergstrom @carlbergstrom.com · 03/10/2026

"The White House and the N.S.F. said that diverting the money was allowed under Congress’s two-year appropriation."

Wenseslio Moreno with talking chicken puppet on the Ed Sullivan Show in 1966.

37118

Carl T. Bergstrom @carlbergstrom.com · 02/10/2026

Your nerd-ass freshman roommate after three beers, repeating random shit he heard on TikTok.

Elon Musk: Unless they figure how to make places actually invisible, "stealth" just means low radar cross section in the 2 to 18 GHz range. Against and advanced enemy, current stealth technology is not worth the design constraints IMO.

3654262

Carl T. Bergstrom @carlbergstrom.com · 28/09/2026

Seriously, guys, just leave Gary home.

Recently, Gary’s buddies requested he not mention Palantir on their nights out together so as to not tank the whole group’s chances of getting laid. “I get nervous now. I don’t bring it up,” he says. (Gary, along with the other daters in this piece, have been given pseudonyms to protect against professional, if not romantic, repercussions.)

610614

Carl T. Bergstrom @carlbergstrom.com · 27/09/2026

Ignore for a moment the question about whether Anthropic trains on user sessions. I am I right in reading this as saying that Anthropic's "never-before-discovered" reverse transcriptase was *patented* three years ago?

Dr. Mestre first encountered the enzymes in jumbo phages in 2022 while searching for new reverse transcriptases. He worked briefly as a consultant for ReNegade Therapeutics, which filed a patent in 2023 on a number of reverse transcriptases, including the ones Dr. Mestre found. He is listed as a co-inventor on the patent filing.

10380112

Carl T. Bergstrom @carlbergstrom.com · 27/09/2026

Google Gemini says that @jevinwest.bsky.social and I aren't to be trusted because in our first lecture, we had a playful example where Jevin presented data the same way that Tim Cook did, and I called bullshit on it. It hallucinates criticism and the only citations are to the lecture itself.

1. The "Explosive Growth" Slide ControversyIronically, the most direct and explicit example of this occurred during their very first "Calling Bullshit" lecture at the University of Washington.The Setup: Jevin West presented a slide to the class showing a graph that purported to show the "explosive, hockey-stick growth" of web traffic to their new Calling Bullshit website. The graph utilized standard marketing axes and compressed timeframes to make the project look like an unmitigated, runaway viral sensation.The "Bullshit": Carl Bergstrom actually had to step in and call "bullshit" on his own co-instructor's slide. When the data was displayed honestly without selective axis truncation or creative smoothing, it showed that while traffic was high initially, it heavily ebbed, flowed, and plateaued over time.The Criticism: While some praised this as a playful, planned pedagogical moment to show that "anyone can fall prey to bad data framing," critics viewed it as a perfect illustration of the double standard: when it serves a personal or professional promotional narrative (like proving your own course is a massive success), the temptation to use flawed, hype-driven metrics is incredibly high.

1227064

Carl T. Bergstrom @carlbergstrom.com · 24/09/2026

Honey that's what we had in school in the '80s.

What’s one word that comes to mind when you hear the name James Talarico?
Jeffrey, 34, independent, Latino, school counselor

Progressive.
Dennis, 62, Republican, white, retired

Neutral.
Jennifer, 50, Republican, white, stay-at-home parent

Same-sex bathrooms in high schools and grade school.

Jennifer, you said “same-sex bathrooms.” Tell me more about why that comes to you.
Jennifer, 50, Republican, white, stay-at-home parent

That’s what he wants to do in schools. I’m totally against that.

1213312

Carl T. Bergstrom @carlbergstrom.com · 24/09/2026

Not that it matters since you just copied their copy of USDA data but the article itself? Pretty sketchy. Check the COIs; I'd prefer a plant protein article NOT funded by the dairy industry. And I'm not an expert in the field but even I know enough to know that this isn't how Venn diagrams work.

Bizarre Venn diagram about the problems with plant protein.

41076

Carl T. Bergstrom @carlbergstrom.com · 24/09/2026

The sourcing is also poor. The data are from the USDA, and were simply reprinted in the article you cited.

4651

Carl T. Bergstrom @carlbergstrom.com · 24/09/2026

Y'all are really phoning it in with this one. Take this graph. It's useless because you didn't bother to describe the denominator. (These are protein amounts per 100g.)

Graph of total protein and essential amino acids for various foods.

1513820

Carl T. Bergstrom @carlbergstrom.com · 23/09/2026

Friends in Adelaide and South Australia more generally: I'm looking at a birding trip (Gluepots/Flinders) in early August 2027. I'd love to swing by the university or anywhere else to give a talk. If you're interesting, drop me a note. cbergst@uw.edu Or just tell me where to go while visiting.

131059

Carl T. Bergstrom @carlbergstrom.com · 21/09/2026

With Gould at least, we don't have to speculate. Here, a 1997 essay of his. www.nybooks.com/articles/199...

Does this argument imply that all genetically and biologically based attributes of our universal human nature must therefore be adaptations? Of course not. Many, if not most, universal behaviors are probably spandrels, often co-opted later in human history for important secondary functions. The human brain is the most complicated device for reasoning and calculating, and for expressing emotion, ever evolved on earth. Natural selection made the human brain big, but most of our mental properties and potentials may be spandrels—that is, nonadaptive side consequences of building a device with such structural complexity. If I put a small computer (no match for a brain) in my factory, my adaptive reasons for so doing (to keep accounts and issue paychecks) represent a tiny subset of what the computer, by virtue of inherent structure, can do (factor-analyze my data on land snails, beat or tie anyone perpetually in tic-tac-toe). In pure numbers, the spandrels overwhelm the adaptations.

The human brain must be bursting with spandrels that are essential to human nature and vital to our self-understanding but that arose as nonadaptations, and are therefore outside the compass of evolutionary psychology, or any other ultra-Darwinian theory. The brain did not enlarge by natural selection so that we would be able to read or write. Even such an eminently functional and universal institution as religion arose largely as a spandrel if we accept Freud’s old and sensible argument that humans invented religious belief largely to accommodate the most terrifying fact that our large brains forced us to acknowledge: the inevitability of personal mortality. We can scarcely argue that the brain got large so that we would know we must die!

1133

Carl T. Bergstrom @carlbergstrom.com · 20/09/2026

Thank you for this. What fun to (re)discover Harold Cazneaux!

A man and two horses with on dusty ground beneath partially clouded sky.

062

Carl T. Bergstrom @carlbergstrom.com · 19/09/2026

Google is like guys, guys....wait up guys! Our AI is dangerous too!

1427439

Carl T. Bergstrom @carlbergstrom.com · 19/09/2026

When we talk about the costs that LLMs impose on society we should not forget the fact that nothing works anymore because everyone is so scared of getting their shit scraped. I can't even use google scholar from my university or starlink because they send too much traffic.

Google
Sorry...
We're sorry...

... but your computer or network may be sending automated queries. To protect our users, we can't process your request right now.
See Google Help for more information.

984461969

Carl T. Bergstrom @carlbergstrom.com · 16/09/2026

The lawsuit challenges the NIH's current policy of using keywords to flag grant applications associated with climate change, DEI, gender, misinformation research, and vaccine hesitancy, keywords that have been used to terminate my own funding as well. Below: just a few of the banned words.

436052

Carl T. Bergstrom @carlbergstrom.com · 14/09/2026

Coming back to fountain pens after a 25 year hiatus, one of the most striking things is the enormous range of inks with new properties I'd never seen before, including shimmer and sheen. But to my great surprise, it turns out that Waterman Purple from the 1990s has a strong gold sheen as well.

Purple ink swatch with gold sheen, "Waterman Encre Violette (1990s)" with Parker vacumatic silver.

61273

Carl T. Bergstrom @carlbergstrom.com · 14/09/2026

Sam Altman is now citing a 10% risk of extinction due to AI in the next decade. Anyone can say any old shit, so put your money where your mouth is, Sam. If you're wrong, in 2036 you pay me a year of your salary. If you're right, I'll have an underwriter pay you 10x that upon proof of extinction.

Reuters: Altman says 10% risk of AI causing extinction `unacceptable'

2126627

Carl T. Bergstrom @carlbergstrom.com · 12/09/2026

Random poaster: So math should start rewarding identification of interesting problems much more. carlbergstrom.com: Hilbert, in jaunty hat, looking at you like wtf.

David Hilbert has no patience for your bullshit

170

Carl T. Bergstrom @carlbergstrom.com · 12/09/2026

New Alfa achievement unlocked: today I finally had a chance to spend a couple of hours driving a 4c, IMO arguably most beautiful piece of automotive design of the past 25 years.

View from the passenger seat. I'm driving an Alfa 4c with the top down.

71434

Carl T. Bergstrom @carlbergstrom.com · 11/09/2026

Please. God answered. Please! God answered. Please! Please! Please? God, Please! God? ...

1911510

Carl T. Bergstrom @carlbergstrom.com · 10/09/2026

shot / chaser, academic services spam edition "Don’t you think that AI can sometimes make a biomedical paper, a humanities paper, and a physics paper sound oddly similar? In research publishing, your voice matters, and so does how you make an argument."

Dear Author,

AI tools have made academic writing simpler. They can help you create polished, fluent drafts in minutes. But fluency is only one part of a strong research paper.

Don’t you think that AI can sometimes make a biomedical paper, a humanities paper, and a physics paper sound oddly similar? In research publishing, your voice matters, and so does how you make an argument.

So, even if you have used AI assistance in your initial draft, Editage’s Premium Editing Service is here to help you restore that discipline-specific voice, precise tone, and language expected by international journals, along with the expertise, human judgment, and nuance that generic AI tools tend to overlook.

2618

Carl T. Bergstrom @carlbergstrom.com · 10/09/2026

I started translating the announcement into plain English, but got too depressed to continue before I even reached the most pernicious parts in the "full letter" linked.

Dear Colleagues,
The U.S. National Science Foundation was established in 1950 to keep American discovery at the frontier. That mission has not changed. But the political landscape has. So it all ends. Now.
We've heard this message clearly from researchers across disciplines and settings. Science today is no longer a straight line from laboratory to application. We aim to cause maximal harm to universities by diverting funds elsewhere. Rather than investing public funding to generate massive positive externalities, we prefer allowing billionaires to set the scientific rese agenda . Someone—you’ll never guess who—convinced us to pour what money we do spend into AI.
To meet this moment, NSF is embracing a portfolio approach, where much more of our funding can be diverted to industry. We are expanding who carries out NSF-supported work and how, to enable ambitious, multidisciplinary inquiry at the scientific frontier, where frontier is defined as what we want to fund. We will find and support exceptional students and early-career researchers, at the stage when research creativity is high, so long as they US citizens of the right kind. Our new Office of Metascience will implement unproven methods to make a show of assessing the effectiveness of new initiatives, and more importantly will leverage that Gold Standard Science bullshit to eliminate programs in geosciences, basic biology, all aspects of social science, and anything else that makes the Christian right uncomfortable.
We’e already pretty much gutted the entire agency. Funding is at a near stand-still, and the agency has become unresponsive to researchers to a degree that was unimaginable only two years ago. We know none of this succeeds without you, NSF's staff, reviewers, awardees, students, and partners across academia and industry. NSF's role is to create conditions for exceptional science to happen more often and more frequently. We are redoubling our commitment to do so through these efforts.

712759

Carl T. Bergstrom @carlbergstrom.com · 10/09/2026

If OpenAI or Anthropic really believe that this technology has a non-trivial probability of destroying humanity within a decade, where is their SCRAM switch?

923330

Carl T. Bergstrom @carlbergstrom.com · 08/09/2026

Anyone remember how Ma Bell used to dox the the whole damn city and deliver it your doorstep free of charge?

1630137

Carl T. Bergstrom @carlbergstrom.com · 08/09/2026

This merlin just flew into my house and landed on my desk while I was taking a zoom call. I now have at least one colleague who thinks I need to come up with better excuses to get out of meetings.

3571165

Carl T. Bergstrom @carlbergstrom.com · 08/09/2026

Michigan painter Jef Bourgeau has come across my feed a few times in the last two days. His website: detroitmona.wixsite.com/artgarage To my untrained eye his snowscapes seem to nod to Lawren S. Harris, who often painted thenorth shore of Superior. First two: Bourgeau; second two: Harris.

Jef Bourgeau painting Looking Back at Sleeping Bear

Jef Bourgeau painting As High As We Could Go

Lawren S. Harris painting The Idea of North

Lawren S. Harris painting: Nerke, Greenland.

717724

Carl T. Bergstrom @carlbergstrom.com · 07/09/2026

Totally off-topic but is this the same owl? Banff area?

1161

Carl T. Bergstrom @carlbergstrom.com · 06/09/2026

Transferring vast sums of money to a so-called "department of war" to do infectious-disease related work doesn't exactly assuage anyone's fears about bioweapons now does it?

The senior NIH official said the new arrangement would untangle civilian public-health initiatives from defence priorities. “We’re trying to get research that’s more Department of War-related back to Department of War,” the official said. “When the public hears that NIAID funds biodefence, they think we’re developing bioweapons, and we’re trying to get out of that business.”

1125055

Carl T. Bergstrom @carlbergstrom.com · 06/09/2026

Talk about boundary violation, emeritscience.com/en/ !

One last message before I leave you alone.
Yesterday I wrote to you about your paper « How publishers can fight misinformation in and about science and medicine. », which we came across while scanning the biomedical literature, and about Charlie, our AI agent for scientific research.
I know you're busy and I respect your time. 10 minutes can be enough to see if Charlie really saves you time.

51006

Carl T. Bergstrom @carlbergstrom.com · 04/09/2026

Life comes at you fast. One minute you're an unknown health economist rushing to publish a hasty study of COVID incidence in Santa Clara and the next thing you know you're signing away a good fraction of the NIH's budget to the Department of Defense.

Bhattacharya post on X.

NIH funds should be used to improve the health of Americans.

Far from siphoning money away from research, this partnership between @NIH and @ DeptofWar focuses on research projects that will drive discoveries to improve the health of American citizens and members of the military.

21600141