#llm bubble

[el-ffej]

Or, to look at the situation from another angle:

Text of posting:

To people who see the performance of DeepSeek and think:

"'China is surpassing the US in AI."

You are reading this wrong. The correct reading is:

"Open source models are surpassing proprietary ones."

DeepSeek has profited from open research and open source (e.g. PyTorch and Llama from Meta). They came up with new ideas and built them on top of other people's work.

Because their work is published and open source, everyone can profit from it. That is the power of open research and open source.

I am very wary of people going "China does it better than America" because most of it is just reactionary rejection of your overlord in favor of his rival, but this story is 1. absolutely legit and 2. way too funny.

US wants to build an AI advantage over China, uses their part in the chip supply chain to cut off China from the high-end chip market.

China's chip manufacturing is famously a decade behind, so they can't advance, right?

They did see it as a problem, but what they then did is get a bunch of Computer Scientists and Junior Programmers fresh out of college and funded their research in DeepSeek. Instead of trying to improve output by buying thousands of Nvidia graphics cards, they tried to build a different kind of model, that allowed them to do what OpenAI does at a tenth of the cost.

Them being young and at a Hedgefund AI research branch and not at established Chinese techgiants seems to be important because chinese corporate culture is apparently full of internal sabotage, so newbies fresh from college being told they have to solve the hardest problems in computing was way more efficient than what usually is done. The result:

American AIs are shook. Nvidia, the only company who actually is making profit cause they are supplying hardware, took a hit. This is just the market being stupid, Nvidia also sells to China. And the worst part for OpenAI. DeepSeek is Open Source.

Anybody can implement deepseek's model, provided they have the hardware. They are totally independent from DeepSeek, as you can run it from your own network. I think you will soon have many more AI companies sprouting out of the ground using this as its base.

What does this mean? AI still costs too much energy to be worth using. The head of the project says so much himself: "there is no commercial use, this is research."

What this does mean is that OpenAI's position is severely challenged: there will soon be a lot more competitors using the DeepSeek model, more people can improve the code, OpenAI will have to ask for much lower prices if it eventually does want to make a profit because a 10 times more efficient opensource rival of equal capability is there.

And with OpenAI or anybody else having lost the ability to get the monopoly on the "market" (if you didn't know, no AI company has ever made a single cent in profit, they all are begging for investment), they probably won't be so attractive for investors anymore. There is a cheaper and equally good alternative now.

AI is still bad for the environment. Dumb companies will still want to push AI on everything. Lazy hacks trying to push AI art and writing to replace real artists will still be around and AI slop will not go away. But one of the main drivers of the AI boom is going to be severely compromised because there is a competitor who isn't in it for immediate commercialization. Instead you will have a more decentralized open source AI field.

Or in short:

[ecrivainsolitaire]

A summary of the Chinese AI situation, for the uninitiated.

These are scores on different tests that are designed to see how accurate a Large Language Model is in different areas of knowledge. As you know, OpenAI is partners with Microsoft, so these are the scores for ChatGPT and Copilot. DeepSeek is the Chinese model that got released a week ago. The rest are open source models, which means everyone is free to use them as they please, including the average Tumblr user. You can run them from the servers of the companies that made them for a subscription, or you can download them to install locally on your own computer. However, the computer requirements so far are so high that only a few people currently have the machines at home required to run it.

Yes, this is why AI uses so much electricity. As with any technology, the early models are highly inefficient. Think how a Ford T needed a long chimney to get rid of a ton of black smoke, which was unused petrol. Over the next hundred years combustion engines have become much more efficient, but they still waste a lot of energy, which is why we need to move towards renewable electricity and sustainable battery technology. But that's a topic for another day.

As you can see from the scores, are around the same accuracy. These tests are in constant evolution as well: as soon as they start becoming obsolete, new ones are released to adjust for a more complicated benchmark. The new models are trained using different machine learning techniques, and in theory, the goal is to make them faster and more efficient so they can operate with less power, much like modern cars use way less energy and produce far less pollution than the Ford T.

However, computing power requirements kept scaling up, so you're either tied to the subscription or forced to pay for a latest gen PC, which is why NVIDIA, AMD, Intel and all the other chip companies were investing hard on much more powerful GPUs and NPUs. For now all we need to know about those is that they're expensive, use a lot of electricity, and are required to operate the bots at superhuman speed (literally, all those clickbait posts about how AI was secretly 150 Indian men in a trenchcoat were nonsense).

Because the chip companies have been working hard on making big, bulky, powerful chips with massive fans that are up to the task, their stock value was skyrocketing, and because of that, everyone started to use AI as a marketing trend. See, marketing people are not smart, and they don't understand computers. Furthermore, marketing people think you're stupid, and because of their biased frame of reference, they think you're two snores short of brain-dead. The entire point of their existence is to turn tall tales into capital. So they don't know or care about what AI is or what it's useful for. They just saw Number Go Up for the AI companies and decided "AI is a magic cow we can milk forever". Sometimes it's not even AI, they just use old software and rebrand it, much like convection ovens became air fryers.

Well, now we're up to date. So what did DepSeek release that did a 9/11 on NVIDIA stock prices and popped the AI bubble?

Oh, I would not want to be an OpenAI investor right now either. A token is basically one Unicode character (it's more complicated than that but you can google that on your own time). That cost means you could input the entire works of Stephen King for under a dollar. Yes, including electricity costs. DeepSeek has jumped from a Ford T to a Subaru in terms of pollution and water use.

The issue here is not only input cost, though; all that data needs to be available live, in the RAM; this is why you need powerful, expensive chips in order to-

Holy shit.

I'm not going to detail all the numbers but I'm going to focus on the chip required: an RTX 3090. This is a gaming GPU that came out as the top of the line, the stuff South Korean LoL players buy…

Or they did, in September 2020. We're currently two generations ahead, on the RTX 5090.

What this is telling all those people who just sold their high-end gaming rig to be able to afford a machine that can run the latest ChatGPT locally, is that the person who bought it from them can run something basically just as powerful on their old one.

Which means that all those GPUs and NPUs that are being made, and all those deals Microsoft signed to have control of the AI market, have just lost a lot of their pulling power.

Well, I mean, the ChatGPT subscription is 20 bucks a month, surely the Chinese are charging a fortune for-

Oh. So it's free for everyone and you can use it or modify it however you want, no subscription, no unpayable electric bill, no handing Microsoft all of your private data, you can just run it on a relatively inexpensive PC. You could probably even run it on a phone in a couple years.

Oh, if only China had massive phone manufacturers that have a foot in the market everywhere except the US because the president had a tantrum eight years ago.

So… yeah, China just destabilised the global economy with a torrent file.

[struungout]

Came across an AI "artist" on Bluesky yesterday where their subjects were BJDs.

[lockandkeyblade]

[gravitasdeficits]

"AI is coming, you have to learn how to use it!"

So is it more complicated than sending it multiple variations of the same question or task as we would search multiple times with different terms in the 90s? Is a "Prompt Engineer" using any kind of shorthand, deeper knowledge, or jargon to get it going better/faster?

[el-ffej]

Regarding the DeepSeek AI Hysteria:

Text of posting:

To people who see the performance of DeepSeek and think: "'China is surpassing the US in AI." You are reading this wrong. The correct reading is: "Open source models are surpassing proprietary ones." DeepSeek has profited from open research and open source (e.g. PyTorch and Llama from Meta). They came up with new ideas and built them on top of other people's work. Because their work is published and open source, everyone can profit from it. That is the power of open research and open source.

Also recommended reading:

[el-ffej]

Brilliantly said. 10/10 no notes.

The problem here isn’t that large language models hallucinate, lie, or misrepresent the world in some way. It’s that they are not designed to represent the world at all; instead, they are designed to convey convincing lines of text. So when they are provided with a database of some sort, they use this, in one way or another, to make their responses more convincing. But they are not in any real way attempting to convey or transmit the information in the database. As Chirag Shah and Emily Bender put it: “Nothing in the design of language models (whose training task is to predict words given context) is actually designed to handle arithmetic, temporal reasoning, etc. To the extent that they sometimes get the right answer to such questions is only because they happened to synthesize relevant strings out of what was in their training data. No reasoning is involved […] Similarly, language models are prone to making stuff up […] because they are not designed to express some underlying set of information in natural language; they are only manipulating the form of language” (Shah & Bender, 2022). These models aren’t designed to transmit information, so we shouldn’t be too surprised when their assertions turn out to be false.

ChatGPT is bullshit

[realityfragments]

A Tale of Two AIs.

2023 has been the year where artificial intelligences went from science fiction to technology possibility. It’s become so ubiquitous that on Christmas Eve, chatting with acquaintances and friends, people from all walks of life were talking about it. I found it disappointing, honestly, because it was pretty clear I was talking about one sort of artificial intelligence where others were talking…

[dvandom]

If you ever want an example of a media bubble that isn't about politics (directly, anyway), this is a damn fine example.

Out in the wild, people either are excited about AI or indifferent. AI-hate is a distinctly minority opinion, and it's mostly "Rise of the Machines" vintage stuff. My workplace is mostly enthusiastic about AI, with the worst attitude being mainly, "Well, it's gonna happen whether we want it or not, so we'd better learn how to use it, and recognize when someone's using it but lying about having done so."

Get on Tumblr or Bluesky or any social media thick with creative types, and suddenly you wonder how anyone could possibly want to use or support the stuff. Clearly it's just something that only techbros want us to use, and even they know it's crap, they're just trying to avoid paying people for art.

THAT is a media bubble. Where you not only get reinforcement of one particular view, but you feel like the opposing views don't really exist to any significant amount.

[canmom]

you read ML research (e.g. arxiv, state of ai, various summaries), you find an overwhelming blizzard of new techniques, clever new applications and combinations of existing techniques, new benchmarks to refine this or that limitation, relentless jumps in capabilities that seem unstoppable (e.g. AI video generation took off way faster than I ever anticipated). at some point you start to see how Károly Zsolnai-Fehér became such a parody of himself!

you read ed zitron & similar writers and you hear about an incomprehensibly unprofitable industry, an obscene last-gasp con from a cancerous, self-cannibalising tech sector that seems poised to take the rest of the system down with it once the investors realise nobody actually cares to pay for AI anything like what it costs to run. and you think, while perhaps he presents the most negative possible read on what the models are capable of, it's hard to disagree with his analysis of the economics.

you read lesswrong & cousins, and everyone's talking about shoggoths wearing masks and the proper interpretation of next-token-prediction as they probe the LLMs for deceptive behaviour with an atmosphere of paranoid but fascinated fervour. or else compile poetic writing with a mystic air as they celebrate a new form of linguistic life. and sooner or later someone will casually say something really offputting about eugenics. they have fiercely latched onto playing with the new AI models, and some users seem to have better models than most of how they do what they do. but their whole deal from day 1 was conjuring wild fantasies about AI gods taking over the world (written in Java of course) and telling you how rational they are for worrying about this. so... y'know.

you talk to an actual LLM and it produces a surprisingly sharp, playful and erudite conversation about philosophy of mind and an equally surprising ability to carry out specific programming tasks and pull up deep cuts, but you have to be constantly on guard against the inherent tendency to bullshit, to keep in mind what the LLM can't do and learn how to elicit the type of response you want and clean up its output. is it worth the trouble? what costs should be borne to see such a brilliant toy, an art piece that mirrors a slice of the human mind?

you think about the news from a few months ago where israel claimed to be using an AI model to select palestinians in gaza to kill with missiles and drones. an obscene form of statswashing, but they'd probably kill about the same number of people, equally at random, regardless. probably more of that to come. the joke of all the 'constitutional AI', 'helpful harmless assistant' stuff is that the same techniques would work equally well to make the model be anything you want. that twat elon musk already made a racist LLM.

one day the present AI summer and corresponding panics will burn out, and all this noise will cohere into a clear picture of what these new ML techniques are actually good for and what they aren't. we'll have a pile of trained models, probably some work on making them smaller and more efficient to run, and our culture will have absorbed their existence and figured out a suitable set of narratives and habits around using them in this or that context. but i'm damned if I know how it will look by then, and what we'll be left with after the bubble.

if i'm gonna spend all this time reading shit on my computer i should get back to umineko lmao

[mostlysignssomeportents]

The Brave Little Toaster

Picks and Shovels is a new, standalone technothriller starring Marty Hench, my two-fisted, hard-fighting, tech-scam-busting forensic accountant. You can pre-order it on my latest Kickstarter, which features a brilliant audiobook read by Wil Wheaton.

The AI bubble is the new crypto bubble: you can tell because the same people are behind it, and they're doing the same thing with AI as they did with crypto – trying desperately to find a use case to cram it into, despite the yawning indifference and outright hostility of the users:

https://pluralistic.net/2023/03/09/autocomplete-worshippers/#the-real-ai-was-the-corporations-that-we-fought-along-the-way

This week on the excellent Trashfuture podcast, the regulars – joined by 404 Media's Jason Koebler – have a hilarious – as in, I was wheezing with laughter! – riff on this year's CES, where companies are demoing home appliances with LLMs built in:

https://www.podbean.com/media/share/pb-hgi6c-179b908

Why would you need a chatbot in your dishwasher? As it turns out, there's a credulous, Poe's-law-grade Forbes article that lays out the (incredibly stupid) case for this (incredibly stupid) idea:

https://www.forbes.com/sites/bernardmarr/2024/03/29/generative-ai-is-coming-to-your-home-appliances/

As the Trashfuturians mapped out this new apex of the AI hype cycle, I found myself thinking of a short story I wrote 15 years ago, satirizing the "Internet of Things" hype we were mired in. It's called "The Brave Little Toaster", and it was published in MIT Tech Review's TRSF anthology in 2011:

http://bestsf.net/trsf-the-best-new-science-fiction-technology-review-2011/

The story was meant to poke fun at the preposterous IoT hype of the day, and I recall thinking that creating a world of talking appliance was the height of Philip K Dickist absurdism. Little did I dream that a decade and a half later, the story would be even more relevant, thanks to AI pump-and-dumpers who sweatily jammed chatbots into kitchen appliances.

So I figured I'd republish The Brave Little Toaster; it's been reprinted here and there since (there's a high school English textbook that included it, along with a bunch of pretty fun exercises for students), and I podcasted it back in the day:

https://ia803103.us.archive.org/35/items/Cory_Doctorow_Podcast_212/Cory_Doctorow_Podcast_212_Brave_Little_Toaster.mp3

A word about the title of this story. It should sound familiar – I nicked it from a brilliant story by Tom Disch that was made into a very weird cartoon:

https://www.youtube.com/watch?v=I8C_JaT8Lvg

My story is one of several I wrote by stealing the titles of other stories and riffing on them; they were very successful, winning several awards, getting widely translated and reprinted, and so on:

https://locusmag.com/2012/05/cory-doctorow-a-prose-by-any-other-name/

All right, on to the story!

One day, Mister Toussaint came home to find an extra 300 euros' worth of groceries on his doorstep. So he called up Miz Rousseau, the grocer, and said, "Why have you sent me all this food? My fridge is already full of delicious things. I don't need this stuff and besides, I can't pay for it."

But Miz Rousseau told him that he had ordered the food. His refrigerator had sent in the list, and she had the signed order to prove it.

Furious, Mister Toussaint confronted his refrigerator. It was mysteriously empty, even though it had been full that morning. Or rather, it was almost empty: there was a single pouch of energy drink sitting on a shelf in the back. He'd gotten it from an enthusiastically smiling young woman on the metro platform the day before. She'd been giving them to everyone.

"Why did you throw away all my food?" he demanded. The refrigerator hummed smugly at him.

"It was spoiled," it said.

#

But the food hadn't been spoiled. Mister Toussaint pored over his refrigerator's diagnostics and logfiles, and soon enough, he had the answer. It was the energy beverage, of course.

"Row, row, row your boat," it sang. "Gently down the stream. Merrily, merrily, merrily, merrily, I'm offgassing ethelyne." Mister Toussaint sniffed the pouch suspiciously.

"No you're not," he said. The label said that the drink was called LOONY GOONY and it promised ONE TRILLION TIMES MORE POWERFUL THAN ESPRESSO!!!!!ONE11! Mister Toussaint began to suspect that the pouch was some kind of stupid Internet of Things prank. He hated those.

He chucked the pouch in the rubbish can and put his new groceries away.

#

The next day, Mister Toussaint came home and discovered that the overflowing rubbish was still sitting in its little bag under the sink. The can had not cycled it through the trapdoor to the chute that ran to the big collection-point at ground level, 104 storeys below.

"Why haven't you emptied yourself?" he demanded. The trashcan told him that toxic substances had to be manually sorted. "What toxic substances?"

So he took out everything in the bin, one piece at a time. You've probably guessed what the trouble was.

"Excuse me if I'm chattery, I do not mean to nattery, but I'm a mercury battery!" LOONY GOONY's singing voice really got on Mister Toussaint's nerves.

"No you're not," Mister Toussaint said.

#

Mister Toussaint tried the microwave. Even the cleverest squeezy-pouch couldn't survive a good nuking. But the microwave wouldn't switch on. "I'm no drink and I'm no meal," LOONY GOONY sang. "I'm a ferrous lump of steel!"

The dishwasher wouldn't wash it ("I don't mean to annoy or chafe, but I'm simply not dishwasher safe!"). The toilet wouldn't flush it ("I don't belong in the bog, because down there I'm sure to clog!"). The windows wouldn't retract their safety screen to let it drop, but that wasn't much of a surprise.

"I hate you," Mister Toussaint said to LOONY GOONY, and he stuck it in his coat pocket. He'd throw it out in a trash-can on the way to work.

#

They arrested Mister Toussaint at the 678th Street station. They were waiting for him on the platform, and they cuffed him just as soon as he stepped off the train. The entire station had been evacuated and the police wore full biohazard containment gear. They'd even shrinkwrapped their machine-guns.

"You'd better wear a breather and you'd better wear a hat, I'm a vial of terrible deadly hazmat," LOONY GOONY sang.

When they released Mister Toussaint the next day, they made him take LOONY GOONY home with him. There were lots more people with LOONY GOONYs to process.

#

Mister Toussaint paid the rush-rush fee that the storage depot charged to send over his container. They forklifted it out of the giant warehouse under the desert and zipped it straight to the cargo-bay in Mister Toussaint's building. He put on old, stupid clothes and clipped some lights to his glasses and started sorting.

Most of the things in container were stupid. He'd been throwing away stupid stuff all his life, because the smart stuff was just so much easier. But then his grandpa had died and they'd cleaned out his little room at the pensioner's ward and he'd just shoved it all in the container and sent it out the desert.

From time to time, he'd thought of the eight cubic meters of stupidity he'd inherited and sighed a put-upon sigh. He'd loved Grandpa, but he wished the old man had used some of the ample spare time from the tail end of his life to replace his junk with stuff that could more gracefully reintegrate with the materials stream.

How inconsiderate!

#

The house chattered enthusiastically at the toaster when he plugged it in, but the toaster said nothing back. It couldn't. It was stupid. Its bread-slots were crusted over with carbon residue and it dribbled crumbs from the ill-fitting tray beneath it. It had been designed and built by cavemen who hadn't ever considered the advantages of networked environments.

It was stupid, but it was brave. It would do anything Mister Toussaint asked it to do.

"It's getting hot and sticky and I'm not playing any games, you'd better get me out before I burst into flames!" LOONY GOONY sang loudly, but the toaster ignored it.

"I don't mean to endanger your abode, but if you don't let me out, I'm going to explode!" The smart appliances chattered nervously at one another, but the brave little toaster said nothing as Mister Toussaint depressed its lever again.

"You'd better get out and save your ass, before I start leaking poison gas!" LOONY GOONY's voice was panicky. Mister Toussaint smiled and depressed the lever.

Just as he did, he thought to check in with the flat's diagnostics. Just in time, too! Its quorum-sensors were redlining as it listened in on the appliances' consternation. Mister Toussaint unplugged the fridge and the microwave and the dishwasher.

The cooker and trash-can were hard-wired, but they didn't represent a quorum.

#

The fire department took away the melted toaster and used their axes to knock huge, vindictive holes in Mister Toussaint's walls. "Just looking for embers," they claimed. But he knew that they were pissed off because there was simply no good excuse for sticking a pouch of independently powered computation and sensors and transmitters into an antique toaster and pushing down the lever until oily, toxic smoke filled the whole 104th floor.

Mister Toussaint's neighbors weren't happy about it either.

But Mister Toussaint didn't mind. It had all been worth it, just to hear LOONY GOONY beg and weep for its life as its edges curled up and blackened.

He argued mightily, but the firefighters refused to let him keep the toaster.

#

If you enjoyed that and would like to read more of my fiction, may I suggest that you pre-order my next novel as a print book, ebook or audiobook, via the Kickstarter I launched yesterday?

https://www.kickstarter.com/projects/doctorow/picks-and-shovels-marty-hench-at-the-dawn-of-enshittification?ref=created_projects

Check out my Kickstarter to pre-order copies of my next novel, Picks and Shovels!

If you'd like an essay-formatted version of this post to read or share, here's a link to it on pluralistic.net, my surveillance-free, ad-free, tracker-free blog:

https://pluralistic.net/2025/01/08/sirius-cybernetics-corporation/#chatterbox

Image: Cryteria (modified) https://commons.wikimedia.org/wiki/File:HAL9000.svg

CC BY 3.0 https://creativecommons.org/licenses/by/3.0/deed.en

[imsobadatnicknames2]

How does one come to the conclusion that generative AI isn't theft if it steals data and takes its training resources nonconsensually?

I don't believe in the concept of intellectual property and I think describing any situation where a copy of something is made and the original version remains unaffected as "theft" or "stealing" is frankly ridiculous.

My opinions on AI itself are far more complex of course, and I generally have very unfavorable things to say about the gen AI tech bubble, if you actually want to hear them in good faith I laid out most of them in this post

Like I said, I have numerous criticisms of the proliferation of generative AI services and the tech bubble surrounding them, but "it's stealing" is not one of them because the idea that making a copy of something can under any circumstances be "stealing" is a belief I'm fundamentally against and you will never catch me legitimizing it.

[nicuveo]

i continue to believe that the only reason we (collectively) get fooled by ChatGPT is that language is easy to mimic, and it fools our primitive brains. LLMs can't reason, can't remember, do not possess any intelligence, but they *mimic* those extremely efficiently.

we, humans, collectively, are primed to trust one another by default (don't believe the naysayers). we are gregarious. so when a machine that talks like a human tells us something in a human-like fashion, of course we trust it by default! if not friend, why friendly?

(sidenote: if anything, it's an indictment of the shallowness of human thought that procedurally generated random sentences can end up tricking us into perceiving a mind, can trick our pattern recognition into hallucinating an intelligence where there isn't. it is no surprise to me that cults are forming around LLMs: we're basically doing a slightly more advanced version of entrails reading: deriving meaning from the hallucinated patterns in random outputs.)

as a result, it is absolutely unsurprising to me that LLMs are bad at chess. they *obviously* are. they do not reason, they do not encode the rules, they purely output statistically likely text based on their training data, that probably does contain some non-negligible amount of chess data; but that's not enough to even play by the rules! even a LLM purely trained on chess books would not be good at chess, because it would still randomly generate illegal moves!

and so, yes: for the very same reason that a LLM is bad at chess, it will be bad for everything else. it cannot remember context. it cannot reason. but language, unlike chess, is more... malleable. rules are not as binary. so, with text, it's less obvious to see when it's garbage, because, again, it's fooling our minds. as that research article eloquently said: it's a bullshitter. and bullshit works.

but, in essence, if you trust a LLM with anything... you're basically taking life advice from InspiroBot.

Deep Blue is 30 years old and was capable of defeating chess grand champions. It could be housed in a single cabinet.

ChatGPT spans untold data centers devouring massive amounts of electricity and it got its ass whipped by an 8 bit gaming console from the 1970s.

[cyle]

still confused how to make any of these LLMs useful to me.

while my daughter was napping, i downloaded lm studio and got a dozen of the most popular open source LLMs running on my PC, and they work great with very low latency, but i can't come up with anything to do with them but make boring toy scripts to do stupid shit.

as a test, i fed deepseek r1, llama 3.2, and mistral-small a big spreadsheet of data we've been collecting about my newborn daughter (all of this locally, not transmitting anything off my computer, because i don't want anybody with that data except, y'know, doctors) to see how it compared with several real doctors' advice and prognoses. all of the LLMs suggestions were between generically correct and hilariously wrong. alarmingly wrong in some cases, but usually ending with the suggestion to "consult a medical professional" -- yeah, duh. pretty much no better than old school unreliable WebMD.

then i tried doing some prompt engineering to punch up some of my writing, and everything ended up sounding like it was written by an LLM. i don't get why anybody wants this. i can tell that LLM feel, and i think a lot of people can now, given the horrible sales emails i get every day that sound like they were "punched up" by an LLM. it's got a stink to it. maybe we'll all get used to it; i bet most non-tech people have no clue.

i may write a small script to try to tag some of my blogs' posts for me, because i'm really bad at doing so, but i have very little faith in the open source vision LLMs' ability to classify images. it'll probably not work how i hope. that still feels like something you gotta pay for to get good results.

all of this keeps making me think of ffmpeg. a super cool, tiny, useful program that is very extensible and great at performing a certain task: transcoding media. it used to be horribly annoying to transcode media, and then ffmpeg came along and made it all stupidly simple overnight, but nobody noticed. there was no industry bubble around it.

LLMs feel like they're competing for a space that ubiquitous and useful that we'll take for granted today like ffmpeg. they just haven't fully grasped and appreciated that smallness yet. there isn't money to be made here.

[wolfliving]

Dead AI IoT

Humane AI finally shuts down, HP pays $116m for the pieces — not including the Ai Pin

By David Gerard on 19 February 2025

Humane, creator of the fabulous and literally nonfunctional Ai Pin gadget — that's "Ai," not "AI" — has finally thrown in the towel.

After Humane took $230 million in venture funding and tried and failed to sell itself for $1 billion, Hewlett-Packard is paying $116 million to acqui-hire most of the team and get Humane’s software and patents. [Humane, archive]

There was also some burbling from HP about “an intelligent ecosystem across all HP devices from AI PCs to smart printers and connected conference rooms,” which probably means floundering a bit then selling the software on once again, as they did with Palm and WebOS.

We’re sure that Humane’s venture capital backers — including Sam Altman, Microsoft, and Marc Benioff — will be delighted that minus-50% is now the expected realized return on AI investments in the bubble.

HP is notably not taking on the Ai Pin itself — probably because it’s completely useless and doesn’t work. The hardware overheats and fails, the projected display isn’t visible in sunlight, and the software chains together LLMs to fail to understand or translate conversations. Also, it might catch fire.

The remaining Ai Pins will work until February 28, when the back-end servers at Humane shut down. After that date, you can ... check the battery level? Pretty good for a $700 gadget with a $24/month subscription. We’re sure both customers will be delighted. ]Humane, archive; Humane, archive]