You are weird

hallucinating plagiarism computer code has feelings too.

for better analysis medium.com/@catesawers/...

📌

Do you ever get tired of being publicly wrong, or is the grift just THAT profitable?

CLANKERS WILL NOT REPLACE US

"wokeness is a coherent political philosophy" well if you were smart you would know it is just an political weapon of the right to just describe leftism as wokeness like being aware of inequality and systemic problems its not coherent political philosophy because it just means improving the world

also you are conflating coherent as dogmatic because far right is dogmatic left is principled they work to make things better for others that is why you cant assign a specific meaning to the word because everyone thinks and acts differently but when you talk to others you find what you have common

wealthy people and right wingers dont talk to strangers that is their uniting force they hate strangers and devalue them their empathy is limited to a small circle some men only feel empathy towards other men and abuse their partners because to them they are not partners in any meaning of the word

Fuk ur mama

I still remember an IBM science rep visiting our engineering school music department (1980s) to demonstrate and play some classical music “composed” by computer. Even we could tell it was “off”. Now seeing the same sort of thing from chat“AI”. The real story seems to be why people fall for it.

And yes, I did try out ChatGPT when it was released. AS SOON as questions/prompts got to a certain level of detail it completely fell apart. Illegal chess moves suggested in discussing a particular opening. Three essentially identical (wrong) quotes when asked, by a TV character in an episode.

And of course it confidently assured me in advance that it could answer questions about precisely these topics. (I asked to “make sure”. Heh.)

In denial or we fucking hate it and would like to see it disappear along with everyone who thinks it's Jesus

If your products actually did any of the crap you claim, you guys wouldn’t have to spend all day every day convincing people. It is the reddest of red flags.

Guy who was wrong about the end of history says we should all really get behind this technology that could end history.

The persistence of corporations using ai as a front to monopolize work, steal intellectual property and data, and use it to spy on us. Fuck ai and fuck you.

If it works so well, why is he devoting so much time to “denialists,” why not just show how well it works?

slurp up day-old dog piss, you fuckwitted charlatan

Dumbarse corporate bootlicker, how’d that “end of history” shit work out again?

Caustic...

Justifiably so

of course AI is fake. history ended 35 years ago didn't you get the memo

is yascha still carrying water for the coup in bolivia?

lmao are you still here

This was obviously written by AI

great another libcon with a blog on nazi substack

Fuck AI and fuck you.

Maybe if you weren't so dumb you wouldn't be tricked so easily by how worthless AI is.

Why don’t you take another vacation from history Francis?

Hey man, history has ended so I ain’t reading this shit.

lol, clown shit

The Peculiar Persistence of the Wallet Inspector Denialists

Just like NFTs revolutionising the financial market. So many people regretting not getting in on those.

Slight problem with AI being this revolutionary world changing technology is that apart from diluting creative spaces with mountains of barely passable slop it has few practical uses, none of which are consumer facing and do not scale with the resources it uses and the environmental damage it causes

Yeah man, we should probably listen to your opinion since you haven’t famously been famously totally wrong about most things before

What people are shipping *is not AI* , and everyone that is using that term needs to be poked in the eye until they stop. Generative models? LLM's God, I hate it when marketing oversells bullshit, media just blindly repeats it, and then people repeat media, and bullshit term becomes the norm.

Thanks for posting this. It sucks, of course, very very bad, but it alerted me to this very good piece the author considered to be bad because of his limited imagination, which was in itself a theme throughout the piece.

The history of the world will be split into a pre-Esperanto and a post-Esperanto era.

log off, frank

Even the end of history repeats itself

Yascha mounk is an idiot

The "AI denialists" are the people who are *still* somehow trying to push this garbage into every corner of existence, despite its drastic flaws. Nobody wants your scam. Stochastic mimicry is not intelligence. It cannot reason. It cannot create. It's an excuse to gut industries for short-term gains

"stochastic mimicry" is such a good way to describe LLMs

Even people who do want their scam don’t really want it, for example only about 10% of ChatGPT’s 100 million users have paid subscriptions.

And we know that about one in 20 users of Anthropic are abusing the system, that is why they are putting rate limits in because these paying users are using up so much it is causing problems. (Also points out that this stuff is likely not profitable).

Ed Zitron’s pointed out the many, MANY ways this stuff isn’t realistically scalable, at least not with how they’re currently going about it.

I have subscribed to zero LLMs, and it leaves me curious how they continue to be funded ad-free. At what point are VC firms going to stop shelling out cash for these things?

That’s the big question. Companies like OpenAI are trying to disrupt a market that doesn’t actually exist, that’s why the comparisons to Amazon and Uber ring so hollow. Bookstores and car services already existed, but there’s no “AI marketplace” because it has no real consumer value.

The actual use-cases for generative AI are things like making flight simulator and surgical training software more realistically unpredictable. The kind of stuff that’s too niche and unsexy to wrap a FOMO sales pitch around.

The fundamental problem is that they invested way too much into a technology that doesn’t have an inherent desirability. I’ve got enough bullshit in my life, thank you.

That’s what I meant by no consumer value. Amazon and Uber were bound to make money eventually because people are always going to need books and car rides, but what material benefit does AI have for average schmoes like you and me?

Exactly. It’s a technology in search of a market. Desperately in search of — they way overpromised.

Then they got some companies on board with forcing it on workers and forcing workers to pretend it’s beneficial while every study shows it slows professionals in their work, causes more errors, and is generally just a waste of money. LLMs are bad business. 😬

Yep! Should have stayed as a graduate student thesis for another 20 years!

I can’t imagine shelling out money for that gimmick

I use ChatGPT, almost exclusively for coding. The free product does what I want to, why would I pay for it? Probably less than 10% of YouTube users pay for premium. Is that evidence that most people who use YouTube don't really want it?

YouTube earns more from ad revenue than they do subscription fees, whereas ChatGPT only generates revenue from subscriptions. Free YT viewers make them millions of dollars, free ChatGPT users cost them millions of dollars.

If subscribers alone were enough to keep the lights on, streaming services wouldn’t have introduced ad-supported tiers and “ad-free” tiers that still make you watch ads.

Yeah, OpenAI is still in the "spend more money than we make" phase to grow the business, just like those streaming services were. It's unknown if OpenAI can pivot away from that. The fact that they haven't yet isn't evidence that they won't be able to. My Netflix sub didn't survive that transition.

The streaming services were making money before they started introducing ads, for a long time the only commercials on any streaming service were on Hulu’s free tier. They started pushing ads because too many people were balking at their rapidly inflating subscription prices.

And that’s where scale comes into play. OpenAI has already set fire to tens of billions of dollars and the cost is only going up, which means they’ll eventually have to go subscription-only. But recent polling shows the vast majority of ChatGPT users will just stop using it if there’s a fee.

They’ve currently got about 10 million users each giving them $200 a year and they’re still nowhere near their break-even point, which itself is just getting farther out of reach. The question is, will they figure out how to keep the lights on once the VC money inevitably dries up?

Yeah, free YouTube users pay for it with their time watching ads. There is no ad-supported chatGPT. But that also goes back to my first point. Why would I pay for it if it does what I want? (1/2)

I just don't think there's enough data to determine if most ChatGPT users would continue using it if there were only ad-supported and paid options. I could probably get my work to pay for the $20 a month plan if there were no free plan. (2/2)

If it can do things that were previously only possible by reasoning or creating, it matters very little whether you think what it's doing technically counts as reasoning or creating. It's like making fun of cars because they don't have legs or feet, and thus can't really run.

If the sun comes up in the west tomorrow, I'll snap my fingers and create three trillion dollars worth of precious metal. Cars work. That's why they're not a scam.

They also worked in 1890, but they were unreliable and had a bunch of shortcomings that lead luddites to dismiss them. There are plenty of reasons to be worried about the impact of LLMs on society, just like the impact of cars. But to dismiss them as a novelty is just dumb.

LLMS have some limited uses in modeling (they're models) but the chatbots touted as "artificial intelligence" are a dead-end technology. The potential for what they're being sold as isn't there, and tellingly, the only improvements being attempted is in the area of making the fakery more convincing.

If in 1890, all the efforts on improving cars was pumped into faking them going anywhere or making them seem steerable when they weren't, you'd have a point. But you don't.

Burro

We're not in denial, you shill. We JUST DON'T LIKE IT. We don't LIKE being stolen from. We don't LIKE the planet being burned down to fuel your auto-complete obsession. We don't LIKE the misuse and endless drive to get us addicted to your perpetual money-making motion engine.

Fuck ur mama too dipshit

Good luck on maintaining your flawless track record of prognostication, mate.

lol how embarrassing for you

You were wrong about "the end of history", and you're definitely wrong about this, lmao.

Ai peddling fuckwits are the ones in denial

Most wrong guy in history has some thoughts in AI.

You're still around? Bummer.

Thank you for confirming my understanding that the current state of AI is 90% hype and 10% "not actually coming anytime soon."

Nobody's ever had to convince me to hop on a technology that actually works. And the spicy autocomplete that losers are hyping into AI isn't it.

“Spicy” as in now autocorrect and autocomplete are both absolutely shit

It's like, you never see commercials for bicycles, because people don't need to be convinced what they're good for.

You’re right. The features of AI that “just work” need little advertisement. Machine translation has been AI for 10 years, and LLMs are better at it*, but you don’t get ads for “chatgpt translates better” Everyone that translates a lot knows that chatgpt is relatively decent for many languages

* for most language pairs. I heard LLMs struggle with eg Norwegian more than DeepL, but most language pairs I do LLMs are better at translating that DeepL and/or Google Translate

i'm pretty sure google translate at least uses an LLM in some capacity now. as someone who has been habitually checking their work learning to write Japanese by machine-translating back to English, i've definitely noticed changes in google translate's output over the last 2 years

as a programmer i'm not convinced about using them for code tbh (and don't bother trying to change that)

Understandable. I have seen that non-programmers can create things that approximate programming. I have not observed a skilled programmer suddenly gaining any new capabilities with ai

Hm, I think an understandable metaphor is like, image generation: a LLM can help programming in the same way it can help you create images: If you’re very bad, it can give you something that kinda maybe looks right? And if you’re good, it might be faster if you need a sketch, though worse.

although honestly both being bilingual (English/Dutch) and learning a third language has made me feel like we've reached diminishing returns on the basic machine translation model and LLMs do better on some metrics and worse than others

i.e. idiomatic translations require a lot of context and i get the impression LLMs are much more likely to make it up, where the old methods were more likely to produce more clunky sentences that don't require it as much

If you translate Japanese regularly, then the differences should be quite visible - that's a language where I see the biggest returns from LLMs. With deepl, you often get it guessing pronouns, and it seems to work line-by-line (and so does google translate) But LLMs seem to be more aware of this

That said, of course they're not perfect and get things wrong, but they're definitely an improvement. Hallucinations are an issue, but mostly a subtle ones - and ones that affect google translate / deepl as well I have one example, evacuation notice that a friend got because flood

And it said people should evacuate from "2:00" GT translated it as "2pm", which is wrong, it was 2am in context. DeepL went for the conservative 2:00, and ChatGPT got 2am presumably from context (it guessed that consistently) I showed it to the (human) international coordinator for the prefecture

(she's really good at Japanese, as you can imagine), and she translated it at 2pm as well - she made the same mistake. It was 2am, and she was like "oh, of course, context" Anyway, I compared all three in more detail, and ChatGPT was the best. Of course, human would be better But yeah, point is:

It's a calculator, but for writing emails. And also it's worse than that because at least a calculator will never give you the wrong answer to 2+2.

You could argue, then, that it's not a calculator at all. :P If someone hands you a device promising to crunch numbers for you and make your day 10x more efficient, and like 20% of the calculations come out with a nonsense result, you wouldn't call it a calculator, you'd call it a waste of time.

Good point. Imagine having to check every calculation a calculator gave you by hand after to make sure it was correct

There are many mathematical problems where verifying the solution is much easier than calculating the solution. If the calculator is wrong 50% of the time, but verifying a solution only takes 10% as much effort as hand calculating the answer, the flawed calculator can still be a useful tool.

Yes, I was just pointing out that the analogy doesn't work when you compare LLMs to calculators specifically. At least to the general image of a calculator I have in my head.

Sure but that's not the promise of what a calculator does

absolutely! I always say: the more they put into marketing of something, the less I wanna buy it! If it was a good product, they'd only have to make ads for it when they initially release it to tell everyone that that's a thing now. If it needs marketing afterwards, its useless

It's not just marketing. They're having to *force* people to use it. That doesn't sound like a technology that does something useful.

Every fast food restaurant ever

fast food ads r da only kind dat werk on me cuz dey trigger instant craving lol

Why are you like this milf (I forgot your actual name 100 years ago but was gonna go for a dramatic sigh and shoulder pat)

You're wrong a lot huh?

Pathetic

i'm going to keep on denying that generative ai is "the future" because my stance is grounded in research, facts and reality. the other side has the promises of techbro billionaires and the lying plagiarism machine. docs.google.com/document/d/1...

📌

Damn straight

📌

📌

Vtnc seu fudido

Ah, the world-class sniffer-of-his-own-farts speaks. How does it feel to be the Oracle of Dumbphi for right-wing pedophiles? P.S your alma mater just rolled over and played dead for Mangolini.

It's cool how all the history kept coming after you said it was ending, huh? Given your opinions are the gold standard for wrongness, I'm glad you're getting on this train, too.

Makes sense the end of history guy would be an ai cultist, since that's also just the rapture but for computers

I thought history ended, Francis. Are you admitting that you're wrong about your most famous quote? Also, AI is literally useless for everything I've used it for, except making writing so limp, uninspiring and corporate that HR drones give it the thumbs up.

he admitted that years ago, update your weapons

Nope. :-P

Sorry, I just realised that "limp, uninspiring and corporate" is a tautology. I could've just said, "corporate".

“The history of the world will be…” I thought that ended back in the 90’s

This is literally the plot of The Terminator. "The history of the world will be split into a pre-AI and a post-AI era."

Mr. Fukuyama, you're a dumb bitch. Ask yourself what your ancestors, who truly would despise your vapid weak ass, have done if they had embarrassed themselves a quarter as much as you have over the years.

Nice to see you out there still being wrong about everything. I'm truly enjoying the End of History

How much money did you lose on NFTs?

Oh, Fran off Fuckcis

You're stupid and you should feel bad about how stupid you are.

hey man, how's post-history going? ya freakin dingbat

[citation needed]

Dumb as shit. Idiot.

wait how is there "a new industrial revolution." that sounds like a major historical event, and i thought we don't have those anymore.

I’m ashamed to have pretended to pay attention to a talk you gave in San Diego.

If you of all people is so confident about AI makes me 100% certain it will fail soon

Hey man, has history already ended?

best to pair any reading of Dr. Fukuyama’s audacious thesis with a soundtrack providing the appropriate gravitas youtu.be/5UoUfGImIUc

you can take Dr. Fukuyama’s opinions as founded in fact and sound reasoning. there's certainly no way his association with Stanford would bring him in contact with "AI" hype cycle grifters or that he'd take any of their outlandish claims at face value and naively parrot them for a wonkier audience

it's so weird how many ppl hate a thing that sucks and is bad

AI managed to make a computer be bad at arithmatic.

People who know nothing about how AI works tend to say things like this, as well as the marketing departments of AI companies

respectfully, Dr. Fukuyama: L

Sure, when we actually invent something intelligent. The lie engine they're currently hyping ain't it though.

stupid man with stupid brain

"The history of the world will be split into a pre-Quibi and post-Quibi era." That's you. That's what you sound like.

"Once, men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them" Frank Herbert, Dune 1965

Could an AI make this?! Didn’t think so…

It can if it stands for Ass Intestines

lol

Me, looking at a board covered with esoteric math formulas: yeah, that checks out.

lol

😂😭

You've been wrong before. Why not be wrong again?

How are you not too embarrassed to ever speak in public again?

You'd think after being so astonishingly wrong all the time, you'd be ashamed to keep showing up and making an ass of yourself, but no, you seem to really get off on it.

i thought history was over

Insanely funny that you stuck yourself to the idea of the end of history, and now you're pivoting to "uh, now everything is new, which is actually what I said all along". You're an irrelevant clown who knows nothing and believes chatbots think you're handsome.

Oh, and also this: softwarecrisis.dev/letters/llme...

Ah! A fellow Bjarnason enjoyer! 😀

“Clap harder!” …No thanks?

Frank, my dude, get it together

Fascinating take from the "might makes right" guy

Oh my god shut the fuck up already

oh, what's that? another "end of history" you say? it'll definitely stick this time

Shut up dumbass

The end of history starts Now. No, I mean… NOW. No, actually, the end is now! This is the real end starting… now!

1) It's not "A.I." just because some VC tech bros call it that. It's just the same algorithms they've been using to come up with Netflix shows since 2012, only now they have 2 billion GPUs and every stolen piece of intellectual property they could shove into it. 2) No one thinks you're smart.

Ai isn't equivalent to fire the printing press or the Internet. You are an clown

The peculiar persistence of fuckheads who think this dumb ai shit doesn't suck rotten ass.

[fart noise]

you're the genius who said history was basically over & fascism was solved, yes?

Perhaps by solved he meant everywhere in hiding

My experience with generative AI so far is that every thing it's cited as being good at by, let's say non-experts in that field, the experts say "Oh, dear god no, it's terrible at this." Even in places where it was believed by users to save time, actual studies have shown it does the opposite.

📌

AI should be very good at checking if a quote is actually in a reference, yet produces work with phantom quotes and phantom references.

"Francis Fukuyama Attempts to Answer the Question 'What is the Most Wrong a Motherfucker Can Ever Be?'"

Hey dipshit how'd that end up history thing end up?

end of historeez nutz

area man seeks new thesis

"Why are so many people resisting this broken piece of shit that lies to people"

Have you considered not being a dumbass?

It turns out that “the more enthusiastic someone is about LLM chatbots, the less they know about how LLM chatbots work” is a pretty reliable rule of thumb.

he probably wouldn't be able to code a chatbot to save his life. lol

The thing about smart people is that they don't tend to stay that way. I'm sorry age has robbed you of something so precious.

Artificial Intelligence is fucking stupid and it will die when the power grids collapse.

I agree with the author. Especially with "I am not a computer programmer, so I don’t have all that many useful things to say about the technology."

Shut the fuck up, goofy.

But i thought history ended 30 years ago, francis! how could new things be happening?

Hi, I'm a software engineer. Your arguments are specious at best, likely disingenuous, and flat-out wrong at worst. I'd go into detail but I'm not going to offer detailed critiques for someone using a Nazi platform.

This sure sounds like some new history Franky

Hey man, for what it's worth, I often refer to The End of History as the most wrong a theoretically smart person has even been in print

No you don't get, ai is the new end of history!! lmao

The history of the world will be split into a pre-AI-bubble-pop and a post-AI-bubble-pop era. Too many people are still in denial.

But hey, didn’t history end like back in…wait a minute

It's that crow at open mic night meme and he's getting booed but all the cards read "end of history"

Ah, you're here. How disappointing.

Francis babygirl what is u doing

The peculiar persistence of the people who are entirely correct

😀🍽️💩🖕

it's perhaps notable that this article does not contain a single concrete example of "Here is a task I had. Here is how I used AI to solve it, and the concrete specific way it helped". The tech is useful in many ways, but that doesn't always align with the hype.

I'll use this essay as an example of the Californian Ideology in my classes www.comune.torino.it/gioart/big/b...

Retire Framcis

in the end he will become the history he so despises

Famously wrong dude wants to be famously wrong again. Chase that fame.

oh jeez, you again

Go back to ending history, bitch

You couldn't be less correct if you tried.

extrapolating current AI capabilities into the future:

Your analysis of the present and future is unparalleled, sir.

No one is denying that AI will change things. The "denialism" is rooted in two facts that the AI Tech Bro Hype Brigade doesn't like. 1. AI is ten or twenty years away from fulfilling any of wild promises made for it. 2. AI is a tool which is used to do harm more than to do good.

Time to disappear old man

Oh god, him

Omg I had no idea this discredited loser was still out there peddling the same kind of bullshit he was spewing in 1993 or whatever

It was still required reading in my poli sci grad program circa 2012! Maybe if fewer people had believed it, we wouldn't have handwaved away the last 10 years of rising fascism, comfortably believing things would never REALLY change. But what do I know.

Perfect!

I may be wrong, but of all people Margaret Thatcher quipped ‘The End of history”, the beginning of nonsense’

This is amazing 😂

Did Fukuyama write this himself or did an AI do it? If the first, he's a hypocrite. If the second, he's useless.

declaring your lack of understanding doesn't convince anyone that you're right it's indicative of an epistemic bubble interfering with the capacity to understand other people's perspectives that's a you problem

pieces like this serve primarily as in group reinforcement for true believers they don't convince anyone to change their mind

Oh, everyone, look! It’s the Wrong Guy, the guy who’s wrong about goddamn everything! He’s doing the thing, being loudly wrong again, clap for him!

Somebody needs to write a paper about how AI's main capability is convincing people of AI's general capabilities, and why so many of those people tend to be anti-woke centrists. Is this just another form of the fetish to identify the final boss of "objective reality" and become its priest?

Alguien lee a este salame, todavía?

LOL

The creators of LLMs convinced tech companies that they could accomplish anything. Tech companies spent billions to buy LLMs and incorporate them into their products. As the limitations of LLMs became clearer, tech companies have scrambled to convince everyone they’re invaluable to save face.

Lol

Occasionally it is worth reading the comments

You can take your AI and you can piss off. Nobody wants what these charlatans are selling. AI is close to its ceiling.

You seem to be operating on the mistaken assumption that AI agents can work. Anyone familiar with information security at an organizational level can explain why they basically can't. Ask yourself why we don't have cities thick with flying cars. Then work sideways.

Thanks for promoting a rapist, Frank. www.thedartmouth.com/article/2024...

didn’t read your stupid ass article, but i can answer a question you’re raising. AI is bad for everyone, and it makes all of our lives worse.

ChatGPT confidently loses chess match to 4K Atari game www.msn.com/en-us/news/t...

Yascha kind of elides Chat's chess skills by focusing on Deep Blue's. Moves that "feel right" don't always work out.

Deep blue- or more relevantly, Stockfish- win at chess because chess is more or less computationally solved. ChatGPT is incapable of doing computations. It’s like that article that claimed ai was old hat in video games because it’s how the ghosts move in pac-man, just an astonishing conflation

This is almost universal among chatbot boosters. Divert from the failures of LLMs by pointing to some completely different computer program that actually works and saying "these are the same!"

Why they also always use AI to mean whatever is most useful in the moment and don't refer to all the different techniques. Or why they seem to sometimes also imagine robots doing additional work which computers cannot do but in an unspoken manner.

I always thought that was what "AGI" originally was really about. i.e., deep blue/alphago are "superhuman" but "narrow" so "generality" is the missing ingredient. take 2 cups alphago, add 1500 mL "generality" and you've got your singularity

Yeah AGI is supposed to be generally intelligent. Btw for the singularity you need extra steps, this AGI needs to be able to improve itself, and this improvement should lead to higher intelligence, and the time between these improvements needs to decrease between steps. Lot of iffy assumptions.

So agi builds agi+ in a week which builds agi++ in four days which builds agi+++ in one etc. Which also means it all must be software, or it needs magic to build new components quickly enough.

But you already see right now with the LLMs that development is slowing down, not increasing, every next step seems to take exponentially more resources. So they are trying to dazzle people with putting addons onto the LLMs which are slight improvements, but not massively new capabilities. Adding a

speech module, a math module. A bit like promising speeds of cars will continue to speed up higher and higher, and then the new model has better tires so it doesn't have to slow down in corners and a radio.

It's worse. His, er, rationale is: Deep Blue beats Kasparov and chatGPT works more like the human mind (according to him), ergo, chatGPT will be better at chess than Deep Blue. And not one exemple to back this up! It's really just faith in the hype (ie capitalism) for all these Yes Men...

Chatgpt is noticeably shit at chess

Oh cool, so he's just got no idea whatsoever how any of this actually works.

Like chatGPT, he communicates with extraordinary confidence to disguise his lack of actual intelligence 😅

He used “astonishingly” soooo many times

He initially bashes critics with the the "astonishing fact that" they don't use ChatGPT, and at the end admits he doesn't "have all that many useful things to say about the technology" and "[struggles] to identify the best people with whom to talk about these topics"

He's smart enough, he's just a gigantic liar and a fraud. (And a rapist.)

Wow! I didn't know that last part!

If you look through the thread there's a few articles people shared about the accusation against him.

I wonder which AI magnate lunatic dug Ol Franky out of the trash bin of history to print this crap lol

It's extremely sad because Chat GPT doesn't even look at a database of grandmaster's games... it looks at things people have written about chess. So it will tell you a move, sometimes ilegal, and tell you why it's good, even if it's terrible and not doing the thing it says.

I don't often play chess, but when I do, I don't make moves based on a "feeling". I do reason, see what moves are allowed, what my adversary could do after that. I'm quickly limited, but bet Kasparov does the same, only a lot lot better. So no, we do play more like Deep Blue than chatGPT.

I bet he thinks other people are stochastic parrots and he's the only person in the world who actually thinks.

If that’s Francis “End of History” Fukuyama, then he should really get out of the prediction business.

Yeah, chess is an odd thing to focus on, given that LLMs seem to be historically bad at chess. There are new models based on huge datasets that are improving it, but it can't do anything a human in its datasets hasn't already done.

Of course it can - you just call it a hallucination.

It can stochastically make an individual move a human hasn't, but it can't develop a strategy or analyze moves, so it would be as unlikely to innovate as any novice.

GPT-4 with "deep analysis" turned on could literally not win a game of Tic Tac Toe when I placed the 1st X in a corner.

Ok. So what? No one claims its AGI. Given training, it could do fine. A 2 year old loses at tic tac toe too, but we don't assume they never will do better.

arxiv.org/html/2407.07...

there is no THERE there to **be** better. There's no persistence & no *thought* or actual *logic.* Just an uncanny impression of the seeming of it, via text probability. How many billions to "train" a program that cannot natively solve tic-tac-toe?

There are already better models, disproving your first statement. Persistence must be coded to the purpose. Logic is part of thinking models. The billions aren't spent on optimizing for tic tac toe. That would be stupid.

this is like when ED209 wasn't "optimized for stairs." You expect sound medical or legal advice, requiring "analysis" from an engine that can't solve tic tac toe from 1st principles? gtfoh

Well, no one except the CEO of the largest "AI" companies: www.tomsguide.com/ai/chatgpt/s... www.reddit.com/r/singularit...

That doesn't claim v4 is AGI.

Still false. arxiv.org/abs/2501.17186

dynomight.net/more-chess/

Except that it isn't - it's still playing at a level well within the skill level of its training data. That's not "innovation" (as Deep Blue innovated), it's pattern recognition. Uncannily good pattern recognition perhaps, but pattern recognition nonetheless.

You said it couldn't do it. Then you said it still couldn't do it. Now you say it just can't do it well. That's moving the goalposts pretty rapidly.

bsky.app/profile/josh...

You know who else relies on pattern recognition? Us.

The chatbots go at it. Kaggle Game Arena AI Chess Exhibition Tournament. The position is White to play and mate in one.

From chess.com

The pro-"AI" people just don't know what they're talking about. They're almost always talking about LLMs, which are trash. If you want to call Stockfish an AI, which is not unfair, yeah, AI kicks ass at chess. If you're saying LLMs are good at chess, well, that's nonsense.

ON TISHABOV I HAVE TO SEE THIS HERETICAL NONSENSE?

Yascha Mounk: wrong about everything

Not sure exactly how many Rs are in strawberry but yeah, it's probably about to rule humanity. Sure dude.

the scam liker has entered the chat

People deny it because it is inconsistent, unreliable, and obviously being recklessly rushed out by sociopath ghouls who are openly slavering over the idea of mass layoffs. Even if it worked, and it doesn’t, it's implementation means mass unemployment and poverty for most while the rich profit

If anyone is interested in the larger movement this pseudo-AI hype is a part of, feel free to read my essay! bsky.app/profile/will...

Aren't you that guy who said the 1990s were the end of history? Do you get off on being massively wrong?

Are we still in history?

The technology seems too young to be so definitive about this, but I think it will revolutionize society like spellcheck did, once all the bugs are worked out. I really like AI personally, but I think we're seeing some over-promising right now.

Are you still going around saying 'no no I didn't mean history would end I meant something else! I'm still technically correct!' Shit, man, look around. Capitalism is eating itself. It's socialism or barbarism, always has been.

Wait… didn’t history end back in 1992?

Lmao

You're a fraud. One of the most embarrassing people ever to be taken seriously as a public intellectual. Please shut the fuck up and live out your remaining years in obscurity, for your own sake if you won't do it for ours.

When confronted with standard problems LLM AI produces working code 2-87% of the time: spectrum.ieee.org/amp/chatgpt-... And most models actually are pretty bad at chess: time.com/7303017/magn... But if I had shit for brains or was over leveraged in this tech bubble, it would be convincing.😉

The most amazing claim to me is the idea that it is valuable for research. Just try doing a Google search for a subject you know and read the ai summary. They are inherent BS generators. That’s why the models collapse when fed their own output. www.nature.com/articles/s41...

I will concede that the author at least admits a strong emotional bias and a lack of relevant expertise at the end, but how does one write about AI without touching on either: A) Hallucinations, which make it disqualifying for the applications proposed B) Linear loss expansion to growth 🤷‍♂️

Keep swimming upstream, pal. AI sucks because tech billionaires are stupid.

The peculiar persistence of people giving Nazis money via substack stanning for ai.

This is dumb and bad, and you are less for having posted it.

every first semester poli-sci grad student laughs their ass off to End of History and now you’re aiming to be the humiliating joke of the computer science field too?

btw when is history going to end thanks in advance

hey francis did you use chatgpt to write this article y/n

It appears that a nearly universal majority of Bluesky posters disagrees with you and Mounk. fwiw I'm a real software engineer ("computer programmer") and I'm on their side. Read more about how these "AI" chatbots actually work. All the marketing around them is a fraud.

If I got something as wrong as the end of history did, I would simply retire from public life forever

And you've never been wrong before so

Yeah. Before, the internet was amazingly useful, and it was fun to run a small site. After, garbage leads results, and you have to install special software to not have the AI companies basically DDOS you.

👶 🧠

you’ve been a midwit since before i was even born.

Not sure you're the best one to demarcate historical epochs, Frankie.

Oh did history start back up again after it ended

Modern AI is the essence of the tech community: the perverse desire to eliminate skill from all human activity. They did it with computing, Thiel's "don't go to to school, be an entrepreneur" and the whole bootcamp, 12 weeks to a programming career thing.

MOOCS and "you can learn everything from YouTube videos", remember that? How about the 20 year old entrepreneur getting his mother to plan his IPO dinner? But that was back when the tech community had to have some contact with reality, in the form of the stock market.

Now, after a decade and a half of venture capital self-soothing, they're left with the ability to push anything? The people who can't be bothered to pick up a pencil and learn to draw shovel their new "art" on the rest of us.

The people who can neither write nor think push their demented hallucinations as the solution to humanity's problems. You think AI is so great? Wonderful! You have no skills, no knowledge, no understanding. Nothing but money. You're at the head of the replacement list.

Lol. Lmao even.

wrong about everything yet again, dipshit!

Plagiarism MATTERS. Environmental destruction MATTERS. Truth MATTERS. If it's "peculiar" to have a moral stance I'm not willing to compromise on then I'm deeply fucking weird.

One of the stupidest "articles" I've ever read that only further convinces me that all proponents of AI have no clue what they are talking about, or what a coherent train or thought looks like. Truly worthy the Francis-Fukuyama-End-of-History-Stamp-of-Approval. bsky.app/profile/phil...