iiinova — where your digital self keeps evolving

Reposted by robert p. baird

Kieran Healy @kjhealy.co · 28/09/2026

Of course, one should bear in mind that correlation does not equal castration.

4939

robert p. baird @robertpbaird.com · 28/09/2026

I won‘t presume that this helps, but: I’m with you.

020

robert p. baird @robertpbaird.com · 23/09/2026

I'm sorry, I know this is very serious and indicative of all kinds of other horrors, but this quote is hilarious. "I am the code," the degraded great-grandchild of "L'État, c'est moi," feels like the purest quintessence of his pusillanimous authoritarianism.

021

Reposted by robert p. baird

MSVH @matthunte.bsky.social · 23/09/2026

This is a beautiful tribute to Annie Dillard by Bobby Baird.

121

Reposted by robert p. baird

Justin Vogt @jvogt.bsky.social · 22/09/2026

From @robertpbaird.com's appreciation of his friend and mentor Annie Dillard. www.newyorker.com/culture/post...

042

Reposted by robert p. baird

The New Yorker @newyorker.com · 22/09/2026

“It is simple enough, and true enough, to say that Annie was unlike anyone I’d ever met, then or since.” Robert Baird remembers Annie Dillard, who died last week, at 81. www.newyorker.com/culture/post...

newyorker.com

Annie Dillard’s Subject Was the Whole World

“It is simple enough, and true enough, to say that Annie was unlike anyone I’d ever met, then or since.” Robert Baird remembers Annie Dillard, who died last week, at 81.

03111

robert p. baird @robertpbaird.com · 22/09/2026

Watching this reminds me of the year I lived in Kampala in 2010-11. The city, appropriately, was obsessed with potholes. They were awful, enormous. Caused traffic jams, accidents, etc. Though ZM was gone by then, his family was not, and I can’t help wondering if there might be a connection.

000

Reposted by robert p. baird

Dan Davies @dsquareddigest.bsky.social · 21/09/2026

My favourite is that "average" comes from an Arabic word meaning "lost goods", and had been used by sailors to describe the practice of fairly sharing storm losses for ages before mathematicians thought that's a useful calculation

35816

Reposted by robert p. baird

Jason Kint @kint.bsky.social · 17/09/2026

Eye popping. Court just unsealed Summary Judgment motions in New York Times v OpenAI / Microsoft. Easy to see why the AI companies had so many redactions (pink). Their own people wrote the lede for NYT, the "largest theft of labor in human history"... "more and more substitutive..." /1

4253422317

robert p. baird @robertpbaird.com · 16/09/2026

Thanks, John. She meant a lot to me. I'm going to miss her.

020

robert p. baird @robertpbaird.com · 16/09/2026

I really don't know what to say, which is probably good because this isn't the place and anyway she hated the usual pieties. Friendship with her was hardly uncomplicated, but right now all I can think of are the wonders she worked on the page and her astonishing generosity off it. RIP, Annie.

250

robert p. baird @robertpbaird.com · 15/09/2026

"…there is something larger at stake here for the left. The failure to think clearly about the pacing proposal is downstream of the failure to think clearly about AI as a whole." — @bentarnoff.com, singing a welcome song bentarnoff.substack.com/p/on-the-pac...

010

robert p. baird @robertpbaird.com · 15/09/2026

At any rate It’s a mistake, I think, to believe there’s a single disagreement about AI that matters more than the rest. The pacing question is important, the energy question is important, the cognitive atrophy question is important, etc. But maintaining human priority has to be high on the list.

000

robert p. baird @robertpbaird.com · 15/09/2026

(He also left GDM under a cloud, and was accused of bullying and workplace abuse: Sebastian Mallaby’s The Infinity Machine has details.)

100

robert p. baird @robertpbaird.com · 15/09/2026

People will no doubt point out that Microsoft is not a frontier AI provider, which is true, though it partnered with OpenAI early on. But it’s worth noting that the head of MAI is Mustafa Suleyman, who was one of the cofounders of Google DeepMind.

110

robert p. baird @robertpbaird.com · 15/09/2026

Some people building AI, such as Sam Altman, really and truly hope to live forever by uploading themselves to a server. Other people think it’s the fate of humanity to be rendered obsolete by machine intelligence.

100

robert p. baird @robertpbaird.com · 15/09/2026

Maybe you think stuff like this goes without saying. And maybe it should. But I assure you that in certain circles, especially circles where decisions about how to build frontier AI are being made, they do not.

110

robert p. baird @robertpbaird.com · 15/09/2026

People who share these concerns might be interested that today Microsoft AI has released a code of conduct for “humanist AI.” I have not read the whole code, but these principles—starting from the first: “people matter more than AI”—seem to me eminently correct.

110

robert p. baird @robertpbaird.com · 15/09/2026

One concern I have, eg—though I lost my earlier post about it—is what happens when certain kinds of utilitarianism (think shrimp suffering) combined with certain views on AI consciousness and moral patienthood (see below) become a demand to weigh “model welfare” against human benefit.

120

robert p. baird @robertpbaird.com · 15/09/2026

There’s been some agitation recently, esp. among people just tuning into the AI safety discussion, about the preponderance of EA/LessWrong types in AI research, and what that means for the safety and ethics proposals on offer.

110

robert p. baird @robertpbaird.com · 15/09/2026

The March of the Ten Thousand where "im not owned! im not owned!!" takes the place of "thalatta! thalatta!"

000

robert p. baird @robertpbaird.com · 15/09/2026

Bluesky's reckoning with AI is shaping up to be as tortuous and torturous as its belated concession that it was a meaningful infirmity, and not a media conspiracy, that pushed Biden out of the 2024 election. Granted that I was tardy on the former point as well, it's still something to behold.

000

robert p. baird @robertpbaird.com · 14/09/2026

A sentence, in retrospect, written precisely for you.

120

robert p. baird @robertpbaird.com · 14/09/2026

Man do I feel this. Part of the problem is that good autofiction makes itself look easy, in a way good maximalist fiction does not. Also, though, I blame the dumb hype that made autofiction out to be something new, even radical, instead of a marketing term for what (some) fiction has always done.

020

robert p. baird @robertpbaird.com · 13/09/2026

But this isn't what's happening? Amodei et al argue: we're building bridges too fast w/o understanding how they work or how to prevent further collapses. Meanwhile, critics say: you're overhyping the dangers of collapsing bridges as a form of marketing, distraction, and/or anticompetitive collusion.

000

robert p. baird @robertpbaird.com · 13/09/2026

Maybe more unsettling is that the companies don't seem to be training their models explicitly to be good at resolving very complicated math problems, and yet lately they have been able to do that too. As Tao and others have argued, this is not necessarily a good thing! But it is not a mirage.

000

robert p. baird @robertpbaird.com · 13/09/2026

…but not many. And I don't think it's hype that changed the picture. To my mind, it's that the companies decided to train their models to be good at coding, and this worked much better than a lot of people (maybe even the companies themselves) thought it could.

100

robert p. baird @robertpbaird.com · 13/09/2026

Maybe our disagreement lies in the degree to which we think that transformation is purely due to the hype. Two years ago coders were reassuring themselves about the kinds of things that AI could not do, which meant that humans were still necessary. There's a few of those assurances still available…

100

robert p. baird @robertpbaird.com · 13/09/2026

(For clarity: I don't necessarily think AI is going to knock everything off op's list, and definitely not to the same degree. I just think we're nowhere near knowing what the hard limits will turn out to be. That can't happen until we see what focused scaling/training in those fields produces.)

000

robert p. baird @robertpbaird.com · 13/09/2026

That helps, thanks. I take op's point to be that past performance is not indicative of future results, and I broadly share that view. AI is doing at lot of stuff that, even two years ago, people thought AI could never do. But I happily admit no one knows, and it's the big open question right now.

200

robert p. baird @robertpbaird.com · 13/09/2026

Don't want to pick a fight on a Sun. morning, but I am curious about this arg. Is it: a/ an "unrelenting failure" to do the things the skilled laborers used to do or b/ an "unrelenting failure" to preserve a decent life for people who might have been skilled laborers previously. Or both? Or neither?

100

Reposted by robert p. baird

Steve Baird @stevenkbaird.bsky.social · 13/09/2026

The reasoning in this thread about AI strikes me as impeccable. His novel is really good too. (And no, we’re not related, although I have to admit he has a very fine name.)

121

robert p. baird @robertpbaird.com · 13/09/2026

Thank you! I admire your name as well.

010

robert p. baird @robertpbaird.com · 13/09/2026

I don't much like glib human-computer analogies. But I can't help noting that one of the major limitations of current LLMs is that their weights are frozen during training, and can't update with new information. Humans don't have that disadvantage, we're told, but sometimes I'm not so sure!

020

robert p. baird @robertpbaird.com · 13/09/2026

I expect that some of the people running AI companies really are thinking about "pacing the frontier" as an anticompetitive tool. But I also think that a lot of the people who talk in that way (Amodei, notably, but also many employees) are not driven primarily or even secondarily by that motivation.

100

robert p. baird @robertpbaird.com · 13/09/2026

I've learned a lot from Rick Perlstein and Atrios, and I don't dismiss them lightly. But it's only fair to note that this reading of the "pacing the frontier" stuff is also the reading of David Sacks, Jason Calcanis, et al, whom I rate...rather differently. rickperlstein.substack.com/p/why-ai-exe...

rickperlstein.substack.com

Why AI Execs Say AI Will Eat You

An annotated bibliography.

220

robert p. baird @robertpbaird.com · 13/09/2026

A shorter way to say what I'm saying below: I think AI is new in important material and conceptual ways. One consequence of that view is that I'm suspicious of claims that present it as something pre-cooked for comprehension, as though it needn't bother to exist for us to understand it.

120

robert p. baird @robertpbaird.com · 13/09/2026

It feels like something similar is going on here. In an effort to preserve its preferred political diagnoses, the left has to insist that AI is nothing new. Even if you hate AI, maybe especially if you hate AI, I continue to believe that this is a major mistake. /end

010

robert p. baird @robertpbaird.com · 13/09/2026

…social relations could change in modernity, thanks in no small part to technological innovation, many of his latter-day followers are so determined to preserve the letter of his conclusions that they end up acting as though nothing new has really happened since the middle of the 19th century.

110

robert p. baird @robertpbaird.com · 13/09/2026

Not “new” in the sense of the laws of physics changing, or even “new” in the sense of a historical disjunction. But new at least in the way the automobile was new, or the Internet. When I spoke to Tooze he noted that while Marx himself was exquisitely attuned to how quickly…

110

robert p. baird @robertpbaird.com · 13/09/2026

I understand that both Doctorow and Seymour have political and ideological motivations for wanting to do that downplaying. And I obviously get the appeal of cold-water thinking. But I really think it’s a mistake not to let oneself consider that something new is happening here.

110

robert p. baird @robertpbaird.com · 13/09/2026

…that it was just doing what it was designed to do, even if its software executed all of its instructions faithfully. Similarly, you have to be really committed to downplaying the novelty of AI to insist that the OpenAI agents are *just* misprogrammed software.

110

robert p. baird @robertpbaird.com · 13/09/2026

In a trivial sense, that statement is true: computer programs are programs, and even if they have randomness or stochasticity built into their programming they are limited by their design. But if an airplane suffers a software bug and crashes into a mountain, I don’t think we gain much by saying…

110

robert p. baird @robertpbaird.com · 13/09/2026

And while I can agree with Doctorow that it’s good not to treat the huggingface incident as though it was perpetrated by little gremlins escaping containment, I also don’t think we get a good picture of the scenario in saying “the chatbots didn't ’go rogue.’ They did what they were designed to do.”

120

robert p. baird @robertpbaird.com · 13/09/2026

I guess it counts as a kind of progress that some people are moving away from the pure stochastic-parrots line. But Seymour’s line (borrowed from Gary Marcus, whether S knows it or not) that it’s the harnesses, and the people designing the harnesses, doing most of the work is only marginally better.

120