One of my empirical-work econ takes is that I think academia should be *more* open to publishing work based on best-we-can-do weak data, as long as there's really no other way to address the question and that work is very clear about what it can and can't actually say.
Ben Sahlmüller
@bensahlmueller.bsky.social
Mathematician obsessed with social organization and change. Speaks bullshit but prefers not to. Fediverse: @b3n@g0v.social
This piece nicely illustrates how even getting to a monopoly is bad for nearly everybody. It also illustrates that the monopoly is also often really bad at what it’s supposed to do. A thing about AI is that it’s probably quite bad at at least a few things, so… newrepublic.com/article/2136...
There’s a Blue-Collar Hollywood—and the Paramount Merger Will Kill It
Hollywood isn’t all glamour. It’s thousands of line workers. And a Paramount takeover of Warner will make their lives even more precarious.
newrepublic.com
A quite apt description of Trump‘s weaving was that he throws ridiculous stuff around and waits to see whether others run with it (the media, his underlings, …). If nobody does it was just a joke, otherwise it can create its own reality. Interestingly, both sides work best if he’s far out.
social epistemology is democracy's biggest challenge. we all know our own experiences, values, and interests. we do not know truth abt the external world and what sort of actions in that world would advance our values and interests. we depend upon institutions for that, and our institutions are bad.
I wonder whether this is the first time fans of Blockbusters And Only Blockbusters encounters cinephiles‘ evaluations of their favorite work. Maybe the death of the cinema meant that most people had no idea people like the ones on letterboxd exists?
Okay now it's getting real
You have to appreciate the craftsmanship and violent playfulness in each of Emily Wilson’s sentences in her Odysseus review, www.lrb.co.uk/the-paper/v4... People seem to debate whether she’s right and whether she Agrees (with you stranger on the internet). Nonsense, just treat this as a rap battle!
Emily Wilson · An Uncomplicated Man
As they say at the awards shows, I was humbled to learn that Christopher Nolan has read at least the first line of my...
lrb.co.uk
Switch the loudspeaker to politics and this is how half of the internet works.
many such cases
I am always mindful of MLK’s profound observation that the hardest (and most unfair) part of any liberation struggle is that the oppressed are obliged not only to free themselves but also to free their oppressor.
Anthropic's new conceptualization of J-spaces in LLMs is very similar to how my "Intuition Behind how LLMs Work" blog post. (Since my blog post is oriented to a non-technical audience, I didn't derive the math. Wasn't interested in doing the math. Now we have the math.) medium.com/@mark-riedl/...
The Intuition Behind How Large Language Models Work, Part I
Large Language Models (LLMs) are fancy artificial neural networks. But you don’t have time to learn the math or engineering. Unfortunately…
medium.com
I like to look at elections as both competition within the democratic system and contention thereof. It’s thus no surprise that parties find contestants of the system among non-voters. The question is whether the party & media landscape can separate both questions - and want to.
One thing that is presumably old news to pollsters but was news to me is that Reform have a lot of success in getting out people who typically don't vote. Kinda similar to Trump? Love how it turns out the modal non-voter was a racist crank. leftwingofthepossible.substack.com/p/makerfield...
I often feel the problem with slop is the disconnect between attitudes of sloppiness and care. It's not inherently bad to do something sloppily, not everything deserves that we put ourselves into it. But right now genAI allows to do things sloppily that signaled care, eg, writing an essay or letter.
Kind of makes you wonder about the role of Harry Potter fan fiction in the training corpus.
This is an actual line that was added to the official system prompt for Codex for GPT-5.5 by OpenAI. Usually the system prompt is as minimal as possible, so I assume it would otherwise mention goblins a lot. AIs are weird.
the costs of child-rearing are straighforwardly measurable and quantifiable. the benefits of child-rearing are wooly and hand-wavy and entirely nonquantifiable. therefore child-rearing is irrational, as a matter of science. your arguments to the contrary are non-falsifiable.
My own recommendation would be: decarbonize electricity production use antitrust to break up oligopolies permissive IP laws limited, strategic public funding of open-source alternatives
🧵 Paul Cadmus (1904-99) was the first fine artist to openly present a gay vision of the world in his time. His work included urban compositions referencing the old masters, portraits & astonishingly sensuous academic studies [1]
Having a mind that works a bit different means that you have to find hundreds and thousands of small cognitive workarounds day by day. I wonder, when LLMs become perfect complements more often for me, can they at the same time also latch well onto normal minds? Only one group makes business sense.
one way to understand all this is the world’s largest petrostate kneecapping its lowest cost competitors, potentially for years, all in the name of protecting the security of their region.
Seems like the people who didn’t like win-win if sometimes the other guy wins more have no problem with lose-lose as long as they can make sure everybody loses more than they do.
this war is bad bad for Europeans Still not fully sunk in
Oh Europe is so screwed 😞 - both CPI↑ & GDP↓ effect much graver - potential Fed rate hikes squeeze Eur finance through $ exposure - influx of refugees & their impact on far right support - Putin propped up, Ukraine weakened - Eur defense buildout slowed - NATO looks doomed (Putin emboldened) etc.
Software is so much about understanding systems and finding ways to make them simpler and easier to manipulate. What I like about agentic coding is that it’s now easier to do this for smaller systems and thus maybe keep them more complex than it was possible in the past.
I have the feeling the right struggles epistemologically with understanding emergent change. It’s inherent to the idea that „the right“ is partly defined by a preference for authority.
This is a clear sign of contempt for the American people and democracy. It's a categorical refusal to submit his actions to public scrutiny.
So, what does „no domestic surveillance“ mean for us Europeans, Mr Altman?
My goodness.
So, if this is more than social media post, does it mean NVIDIA is not allowed to do business with Anthropic any more? That sounds neither smart nor legal…
Hegseth didn’t seem the guy to use the Oxford comma to me.
How much of AI will be rent-seeking, the same old game? I guess it depends on what practices it finally supports. Genuine R&D? Actual growth. Advertising, lawyer-ing, academic games dislodged from outcomes, entertainment? Rent seeking. AI just as earlier tech just raises the bar in a zero-sum game.
its funny too bc everyone can tell that either AI is gonna pull profits in from elsewhere in the system so minimal GDP gain out of their profitability or it hits everything with The Speedup and gets GDP up thru the Velocity of Circulation channel but not clear why that would happen outside marginals
I saw a live birthed harbor seal pup for the first time in my life today. I thought it would be horrifying because I have a weak stomach, but it turned to UNGODLY CUTE very fast! Using this thread for some moments in the minutes after. All shot on my z8, edited on phone for now. 🦭 🦑
Disturbingly detailed analysis of my first rewatch incoming