In a move that is not in any way ominous, and everyone involved has carefully thought through all the consequences, there’s a sora-generated video of sam altman shoplifting gpus that’s apparently quite popular right now.

https://bsky.app/profile/drewharwell.com/post/3m23ob342h22a

(no embed because safari on ipad is weird about downloading or linking video)

I suspect it is also hiding some rendering artefacts.

Oh hey, bay area techfash enthusing about AI and genocidal authoritarians? Must be a day ending in a Y. Today it is Vercel CEO and next.js dev Guillermo Rauch

https://nitter.net/rauchg/status/1972669025525158031

I also have strong opinions about not using next.js or vercel (and server-side javascript in general is a bit of a car crash) but even if you thought it was great you should probably have a look around for alternatives. Just not ruby on rails, perhaps.

I will say that the flipping between characters in order to disguise the fact that longer clips are impractical to render is a neat trick and fits well into the advert-like design, but rewatching it just really reinforces how much those kids look like something pretending real hard to be a human.

Also, fake old-celluloid-film filter for something that was supposed to be from 20 years ago? Really?

AI video generation use case: hallucinatory RETVRN clips about the good old days, such as, uh, walmart 20 years ago?

It his the uncanny valley triggers quite hard. It’s faintly unsettling t watch at all, but every individual detail is just wrong and dreamlike in a bad way.

Also, weird scenery clipping, just like real kids did back in the day!

https://bsky.app/profile/mugrimm.bsky.social/post/3lzy77zydrc2q

Today’s related news: the tailwind css guy is a big fan of dhh and the rubygems takeover.

https://bsky.app/profile/jaredwhite.indieweb.social.ap.brid.gy/post/3lzofv4wi4yz2

I miss the days when being publicly fashy was considered poor pr, but on the other hand it does make it a lot easier to avoid their companies and products.

Tailwind is pointless, incidentally.

Does ruby just die now?

Part of the background to this issue is the development of rv which apparently offers a future where rubygems is much less important, and some folk seem to be taking that as a threat.

Whether or not the new tooling delivers, the rubygems debacle has probably helped the new project considerably.

Haven’t read the source paper yet (apparently it came out two weeks ago, maybe it already got sneered?) but this might be fun: OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws.

Full of little gems like

Beyond proving hallucinations were inevitable, the OpenAI research revealed that industry evaluation methods actively encouraged the problem. Analysis of popular benchmarks, including GPQA, MMLU-Pro, and SWE-bench, found nine out of 10 major evaluations used binary grading that penalized “I don’t know” responses while rewarding incorrect but confident answers.

I had assumed that the problem was solely technical, that the fundamental design of LLMs meant that they’d always generate bullshit, but it hadn’t occurred to me that the developers actively selected for bullshit generation.

It seems kinda obvious in retrospect… slick bullshit extrusion is very much what is selling “AI” to upper management.

Woke up to some hashtag spam this morning

AI’s Biggest Security Threat May Be Quantum Decryption

which appears to be over of those evolutionary “transitional forms” between grifts.

The sad thing is the underlying point is almost sound (hoarding data puts you at risk of data breaches, and leaking sensitive data might be Very Bad Indeed) but it is wrapped up in so much overhyped nonsense it is barely visible. Naturally, the best and most obvious fix — don’t hoard all that shit in the first place — wasn’t suggested.

(it also appears to be a month-old story, but I guess there’s no reason for mastodon hashtag spammers to be current 🫤)

One to watch from a safe distance: dafdef, an “ai browser” aimed at founders and “UCG creators”, named using the traditional amazon-keysmash naming technique and and following the ai-companies-must-have-a-logo-suggestive-of-an-anus style guide.

Dafdef learns your browsing patterns and suggests what you’d do next After watching you fill out similar forms a few times, Dafdef starts autocompleting them. Apply with your startup to YC, HF0 and A16z without wasting your time.

So… spicy autocomplete.

But that’s not all! Tired of your chatbot being unable to control everything on your iphone, due to irksome security features implemented by those control freaks at apple? There’s a way around that!

Introducing the “ai key”!

A tiny USB-C key that turns your phone into a trusted AI assistant. It sees your screen, acts on your behalf, and remembers — all while staying under your control.

I’m sure you can absolutely trust an ai browser connected to a tool that has nearly full control over your phone to not do anything bad, because prompt injection isn’t a thing, right?

(I say nearly full, because I think Apple Pay requires physical interaction with a phone button or face id, but if dafdef can automate the boring and repetitive parts of using your banking app then having full control of the phone might not matter)

h/t to ian coldwater

New lucidity post: https://ludic.mataroa.blog/blog/contra-ptaceks-terrible-article-on-ai/

The author is entertaining, and if you’ve not read them before their past stuff is worth a look.

It isn’t clear to me at this point that such research will ever be funded in english-speaking places without a significant set of regime changes… no politician or administrator can resist outsourcing their own thinking to llm vendors in exchange for funding. I expect the US educational system will eventually provide a terrible warning to everyone (except the UK, whose government looks at the US and says “oh my god, that’s horrifying. How can we be more like that?”).

I’m probably just feeling unreasonably pessimistic right now, though.

Some people casting their eyes over this monster of a paper have less than positive thoughts about it. I’m not going to try and summarise the summaries here, but the threads aren’t long (and are vastly shorter than the paper) so reading them wouldn’t take long.

Dr. Cat Hicks on mastodon: https://mastodon.social/@grimalkina/114690973548997443

Ashley Juavinett on bluesky: https://bsky.app/profile/analog-ashley.bsky.social/post/3lru5sua3fk25

It is related, inasmuch as it’s all generated from the same prompt and the “answer” will be statistically likely to follow from the “reasoning” text. But it is only likely to follow, which is why you can sometimes see a lot of unrelated or incorrect guff in “reasoning” steps that’s misinterpreted as deliberate lying by ai doomers.

I will confess that I don’t know what shapes the multiple “let me just check” or correction steps you sometimes see. It might just be a response stream that is shaped like self-checking. It is also possible that the response stream is fed through a separate llm session when then pushes its own responses into the context window before the response is finished and sent back to the questioner, but that would boil down to “neural networks pattern matching on each other’s outputs and generating plausible response token streams” rather than any sort of meaningful introspection.

I would expect the actual systems used by the likes of openai to be far more full of hacks and bodges and work-arounds and let’s-pretend prompts that either you or I could imagine.

It’s just more llm output, in the style of “imagine you can reason about the question you’ve just been asked. Explain how you might have come about your answer.” It has no resemblance to how a neural network functions, nor to the output filters the service providers use.

It’s how the ai doomers get themselves into a flap over “deceptive” models… “omg it lied about its train of thought!” because if course it didn’t lie, it just edited a stream of tokens that were statistically similar to something classified as reasoning during training.

deleted by creator

I might be the only person here who thinks that the upcoming quantum bubble has the potential to deliver useful things (but boring useful things, and so harder to build hype on) but stuff like this particularly irritates me:

https://quantumai.google/

Quantum fucking ai? Motherfucker,

• You don’t have ai, you have a chatbot
• You don’t have a quantum computer, you have a tech demo for a single chip
• Even if you had both of those things, you wouldn’t have “quantum ai”
• if you have a very specialist and probably wallet-vaporisingly expensive quantum computer, why the hell would anyone want to glue an idiot chatbot to it, instead of putting it in the hands of competent experts who could actually do useful stuff with it?

Best case scenario here is that this is how one department of Google get money out of the other bits of Google, because the internal bean counters cannot control their fiscal sphincters when someone says “ai” to them.

And back on the subject of builder.ai, there’s a suggestion that it might not have been A Guy Instead, and the whole 700 human engineers thing was a misunderstanding.

https://blog.pragmaticengineer.com/builder-ai-did-not-fake-ai/

I’m not wholly sure I buy the argument, which is roughly

• people from the company are worried that this sort of new will affect their future careers.
• humans in the loop would have exhibited far too high latency, and getting an llm to do it would have been much faster and easier than having humans try to fake it at speed and scale.
• there were over a thousand “external contractors” who were writing loads of code, but that’s not the same as being Guys Instead.

I guess the question then is: if they did have a good genai tool for software dev… where is it? Why wasn’t Microsoft interested in it?

Turns out some Silicon Valley folk are unhappy that a whole load of waymos got torched, fantasised that the cars could just gun down the protesters, and use genai video to bring their fantasies to some vague approximation of “life”

https://xcancel.com/venturetwins/status/1931929828732907882

The author, Justine Moore is an investment partner at a16z. May her future ventures be incendiary and uninsurable.

(via garbageday.email)

I was reading a post by someone trying to make shell scripts with an llm, and at one point the system suggested making a directory called ~ (which is a shorthand for your home directory in a bunch of unix-alikes). When the user pointed out this was bad, the llm recommended remediation using rm -r ~ which would of course delete all your stuff.

So, yeah, don’t let the approximately-correct machine do things by itself, when a single character substitution can destroy all your stuff.

And JFC, being surprised that something called “YOLO” might be bad? What were people expecting? --all-the-red-flags