This message which is ironically just above this one on my feed explains it all.

Hypernormalization of the untenable and then collapse.

bsky.app/profile/70sb...

The steelman is a sense of prioritarianism. Science doesn’t have to compete with welfare but I can’t fault people for the mistaken belief when politicians and businessmen say there isn’t money for X.

The thread is about that… literally the second reply says the bubble is mostly in data centers. The rest of the thread acknowledges LLMs aren’t going away and talks about questions of scope and governance.

What to expect when you’re expecting a (AI) bubble Don’t look to the dot-com bubble to understand the consequences of the current one.

(10/10) While I worry about LLMs as a poorly scoped technology, I don't think they are going away. As naive as this sounds, I want responsible, thoughtful people in spaces where they can push back or emphasize LLM failure modes. misaligned.markets/what-to-expe...

What to expect when you’re expecting a (AI) bubble Don’t look to the dot-com bubble to understand the consequences of the current one.

(9/10) LLMs are a non-neutral form of AI built around the AGI delusion and desire to enclose transformers which were public knowledge. User psychosis and model hallucination aren't incidental. This "good" bubble cost us a better scoped technology. misaligned.markets/what-to-expe...

What to expect when you’re expecting a (AI) bubble Don’t look to the dot-com bubble to understand the consequences of the current one.

(8/10) In my post I argue that governance + material constraints shape a technology. There would have been internet without the dotcom dark fiber bubble. There would have been transformers without GPTs and LLMs. misaligned.markets/what-to-expe...

What to expect when you’re expecting a (AI) bubble Don’t look to the dot-com bubble to understand the consequences of the current one.

(7/10) Something else that the "good" bubble framing misses is that bubbles aren't neutral. There are other things society could have done with the resources that went into a bubble. Our current information landscape is incapable of considering those counterfactuals. This motivated my post.

LLM “intelligence” is a dark pattern LLMs leverage scaffolding and user psychology to appear as slabs of raw intelligence in service of a costly illusion.

(6/10) Discussions about AI automation hinge on the unnamed premise of capital-as-labor: a perfect substitution of human for machine. But if models alone aren't the locus of LLMs' efficacy, then this idea is just wrong. It'll take labor to manage AI.

misaligned.markets/llm-intellig...

AI hype is a mirror of market fundamentalism Both AI enthusiasts and market fundamentalists gloss over the context needed to understand complex systems.

(5/10) GenAI doesn't go away after the data center and AI startup bubbles, but the future looks closer to driverless cars. Basically infrastructure intensive tech that looks impressive because it hides its dependencies from view.

misaligned.markets/ai-hype-mark...

What to expect when you’re expecting a (AI) bubble Don’t look to the dot-com bubble to understand the consequences of the current one.

(4/10) Signs that GenAI is a "good" bubble are at best mixed. Scaling infrastructure (annotators, orchestration, more compute, tooling, workflows) is what is driving value. Costs are falling if you ignore this misaligned.markets/what-to-expe...

(3/10) Somewhat separately, the current crop of GenAI companies are unprofitable, but if that were it I don't think the bubble would be as big a problem. A lot of discussion I've seen about the bubble does fixate on this, though. That's where the notion of a "good" bubble comes from.

What to expect when you’re expecting a (AI) bubble Don’t look to the dot-com bubble to understand the consequences of the current one.

(2/10) The "real" bubble is in borrowing to build data centers (and buy GPUs). In my blog post I screenshot an industry marketing post eagerly writing about the premise of borrowing money to build tenantless data centers.

misaligned.markets/what-to-expe...

What to expect when you’re expecting a (AI) bubble Don’t look to the dot-com bubble to understand the consequences of the current one.

(1/10) I think by now everyone knows that AI (LLMs) is a bubble. Just a day after I published this post #Sora was shut down. The real question that needs to be answered is what type of bubble this is. A thread 🧵:

misaligned.markets/what-to-expe...

#AI #AIbubble #LLMs

I've insisted on using the phrase "capital-as-labor" to discuss claims made about LLMs. That said, I think driverless cars are probably the closest parallel we have. Rather than a one to one sub of capital for labor, you have a whole industry behind the scenes building infra to manage AI models.

Cartoon by @goldingcartoons.bsky.social

I think local models with bespoke workflows will have value. Not that it’ll make the bubble and current social costs worth it. But yeah there will also be space for grifters to keep the grift going, so long as people don’t understand how LLMs work.

I don't know where that leaves us. I've called for technical people to aggressively scope the technology and show off defined use cases. Idk how well that scales. But there's no truly "neutral" version of GenAI/LLMs. Harms go down as they're scoped and people are educated, though.

The risk of GenAI disempowering people doesn't go away even if Big Tech leaves. The tech is a non-neutral implementation of the transformer architecture. LLMs exist not in spite of the AGI delusion, but because of it. So they'll always inherit risky flaws like hallucinations, user psychosis, etc.

But I tend not to like analysis that uses the letters "AI." Term is a marketing buzzword that mixes grokified smut generators with genuine predictive models. Also for GenAI/LLMs specificaly, part of the reason I started writing about them is that they're as likely to disempower as empower.

Reverse centaurs are the answer to the AI paradox Not what the machine does, but who it does it to.

Sounds like Cory Doctorow's Reverse Centaur thesis. I think that's directionally correct in that it promotes agency. Also critically highlights the tech won't disappear after bubble, so we should come up with a positive vision before big tech regroups and reframes the post-bubble narrative.

They protected Elon, SBF, Sam Altman... and their fearmongering over AI justified investment. Definitely not beating the allegations that they exist to launder billionaires' views and reputation. Take that away and there's nothing novel. Charities and consequentialism existed before EA.

Alice Crary · Against ‘Effective Altruism’ (2021) Effective Altruism (EA) is a programme for rationalising charitable giving, positioning individuals to do the ‘most good’ per expenditure of money or time. It was first formulated – by two Oxford phil...

But EA genuinely has a measurement problem. They over-index on easy metrics like QALY and don't engage in contextual/systemic thinking about global issues. Measuring outcomes is hard! Knowing what to measure and why is even harder! Alice Crary has a great article on this.

Even granting a consequentialist view, I'm skeptical that sending a bunch of neurotic young people into high externality industries to pay for local solutions like mosquito nets is the greatest good. At least compared to investing in coordinated/systemic solutions. In the absence of that... fine.

So:

A cartoon showing soldiers sitting in a transport plane with jump door open. One soldier asks "Where are we heading, sarge?". The soldier next to him answers "Not sure. But @DonnieJunior just made a $150M Polymarket bet on Kharg Island beachfront futures. By Wintersart (substack.com@wintersart).

This is all too real. These days, satire is just news but in teaser format.

It can be turned off I think.

Cartoon by @andertoon.bsky.social