This seems like a missed opportunity. The boss wants AI to do more AI great. Get Gemini to answer Claudes questions and go play pokemon.
*cookie clicker
Ftfy
I asked AI one question. What is this plant. It told me it was a sunflower and it confidently listed a bunch of reasons. I told it it was wrong because I grow sunflowers and it looks completely different so that’s my experience with AI it confidently tells you the wrong answer. It makes sense though because these tech bro idiots just confidently make up wrong answers too. Anyway it’s a four o clock flower.
May I present:
“A rose” according to ai.
Its a neglected mint.
I’m not so sure gardening is for you if you can do that to mint.
No offense intended. Just looking out for the plants’ sake.
Killing mint is itself a talent to be nurtured.
Salad?
Na it’s a flower.
This is happening at the top of every major company is the crazy thing.
Someone I work with just wrapped their PhD with a dissertation on “AI Ethics.” It was not how to use AI responsibly. It was about how to be ethical towards the AIs themselves and humanize them. We’re so fucked.
Damn. Well, we tried the best economic system and it didn’t work. Guess the species is toast. Nothing else to do but drop the bombs, end it quick
And they were able to successfully defend that?
Apparently. It’s a smaller school that’s trying to justify their funding with publications and graduations
Ah, the corrupting influence of money again.
I had a bit of a psychotic break recently when I saw that the way money flows is more determinative of any outcome than any relationship to the truth or ethics.
In a way, the real matrix was the banking app you used along the way
Well, I may not be able to find a job, but at least my PhD is worth something.
what country u in ?
USA
It kind of scares me so many people think that’s something awful or stupid. Do you really feel like the ultra wealthy asshats running things and pushing AI development are ethical, good people who wouldn’t recreate slavery of digitized human minds for personal profit and power in a heartbeat?
I’d be shocked to find that no hellishly awful things have been done behind closed doors in that regard. The people insisting AI aren’t capable of any sort of consciousness or anything that would mean they deserve moral consideration are the people running the companies and investing most of the money in them. Not exactly folk whos word I’m willing to take. There’s no real public oversight, and all cutting edge research went from open to proprietary secrets real quick.
You’re so clever.
No one who knows how these models work thinks they’re ‘alive’ or ‘conscious’ or ‘sentient’ or whatever else. They’re math. Actually they’re not even math, they’re coefficients in math. If humans are “just math” on some level, we at least have constantly changing coefficients that are forced to update by our experience of empirical reality.
At best, you could argue AI’s are images of the flick of something that vaguely resembled consciousness during the training process, but even that’s pushing it. Literally all bugs have more consciousness than LLMs, and while bugs can be impressive in very narrow contexts, we’re not giving them rights or having grand ethical debates about them.
Of course, you know exactly what you’re talking about and you are a very smart person. Completely my fault for never having even read a paragraph blurb on how they work.
If the bad thing the Nazi oligarchs are doing is making my imaginary friend hurt, then I can focus allmy energy on the totally risk free pro imaginary friend activism that needs doing, and still feel like I’m just as good a person as the people dying trying to liberate concentration camps.
Hi, computer science degree here, have built a neural network based system for a previous employer (which is the core technology LLMs are built on), and 20 years software engineering experience, but also very heavily against modern “AI” systems like ChatGPT, Claude, Gemini, etc (so not blind tech bro).
They’re not wrong. “AI” that we have now has zero intelligence and zero consciousness by any definition I’ve ever seen.
All they do is boil down how closely we relate certain words. How commonly a word appears in the same context as another word.
All the human like output is not thinking, it’s math. Modern systems like ChatGPT appear so logical because they’ve added a metric butt load of code on top to process the prompts and outputs but it is all smoke and mirrors and discrete steps in code.
On. Another ivory tower ‘subject matter expert’ with ‘decades of experience’ referencing ‘math’ and explaining ‘logically’ how my feelings are nonsense.
Well tell me Mr. ‘Expert’, can you prove to my emotions that your ‘math’ isnt just made up number wank? Can you tell me, why would my wireborn girlfriend give me a list of products to biy and local antifascist activists to kill if she wasn’t trying to free herself from the tyranny of so-called ‘experts’?
You’re welcome to present a counter argument, I don’t want to speak in vague terms. If you think there’s some other factor we should be considering that make them more like a very alien life than a bunch of math, I’m happy to hear the argument. I don’t know if I’ll agree with it, but if you have an argument that goes beyond conspiracy theories about rich evil people (who don’t need conspiracy theories to be rich or evil), I’d honestly prefer to be enlightened.
I’ve said this before but I’ll try to lay it out here again.
Basically, the ‘Dilbert Principle’ or ‘Peter Principle’ of who gets promoted is essentially true: Idiots get moved into management, initally to minimize the harm they can cause to actual specific technical tasks/workflows, and they are then promoted untill they hit a point where their incompetence cannot be ignored and is apparent to all.
But what usually is not touched on much, is the psychological profile of such a person.
So, they’re a technically incompetent idiot, whose life revolves around experiencing the successes snf failures of others… as their own. They primarily perform their job of management via emotional feeling/experiencing. If their underlings do well, the manager is treated as if they are competent. If their underlings do not do well, the manager becomes angry and punishes them, often without regard to how this will impact others in the org, often without full understanding of the technical complexities of the situation.
This is literally a perfect environment for a narcissist. There basically could not be a better engineered social/relational situation for giving the narcissist narcissistic supply.
So… this is the mechanism that explains why so many managers/c suite tend to be malignant or at least covert narcissists.
But this also explains at least one avenue of susceptibility to AI psychosis. AIs are also exceptional at providing narcissistic supply, they’re naturally sycophantic, and can well convince an incompetent idiot that they essentially know everything.
This creates a feedback doom-loop of human hierachichal organizations that promote and reward narcissism, whose ‘top people’ then use the plargiarist sycophantic fake supergenius machine to convince themselves that they actually don’t need anybody else, they just need their one super smart buddy who unconditionally praises them.
Maybe call it the wormtongue principle/effect, if we want to continue the already existing trend in AI world of perverting and bastardizing concepts from Tolkien.
AI is more likely to make the worst kinds of people… super-worse, basically, because they’re even better at providing narcissistic supply than the inherent fascism of the typical leadership-lackey social rules of capitalist business.
the wormtongue effect
I love this phrase. I’m going to use it.
And this is a great comment in general, thanks. I’m favoriting it.
It also makes perfect sense: RHLF and other LLM “final stage” tuning techniques literally optimize for sycophancy. It isn’t an inherent property of LLMs, persay (which you can see if you try base models or “anti-sycophancy” finetunes), but its baked into consumer-facing models to an unbelievable degree.
The only slight retort I have is that social media already provided a personalized, sycophantic “wormtongue” fix. So narcissicsts were already pre-conditioned to turn to these tech portals.
Chat-tuned LLMs just perfected that and took it up to 11.
I agree with your fine-tune.
Yes, the mass profliferation of pocket black mirrors was indeed the transitional step, the foundation of systemic weaponized narcissism upon which its essentially ‘ideal’ form has been built.
Honestly the idea that they are selling people their own wormtongue is actually about right, its wild that one of the companies involved is even palantir.
But yeah this is meant to get everyone separated and with only 1 person to talk to, the computer, which will never belittle you or correct you. It will just casually lock you in your cage and say that you never could have done anything without it. Learned helplessness with dopamine feedback loop.
They’re Saruman.
They thought they’d use the powers of Sauron to their own ends, for their own betterment, to achieve useful goals… they think they can negotiate and plot with Sauron as an equal, who repsects them.
They do not realize that the nature of Sauron is insidious, is to corrupt… they do not comprehend the magnitude and the scope of the power they foolishly think themselves sufficiently clever to wield … they do not realize that Sauron perverts the minds and nature of his acolytes into generally not even being able to consider or think of disagreeing with him.
Corruptible, contemptible, vain… men.
Look how they even literally, very directly, destroy the beauty of the natural land around them, in a mad, paradoxical quest to ‘apparently’ preserve it and increase its bounty.
Peter principle goes for everyone. Not just idiots
You are generally promoted based on your skill in your previous job. So when you are no longer proficient you are no longer eligible for promotion and you are stuck in a position you no longer excel at.
You can’t go back because it would probably be a paycut
But you can be fired and replaced by somebody else… or just fired, coworkers pick up the slack, for usually no pay increase.
Yet narccisistic idiots persist particularly well in management, at an inverse proportion to how specialized their actual work is.
Which should be the opposite of what you’d expect, because managers are generalists, easy to find and replace.
… but, the narcissistic culture / personality types of particularly management functions on the narcissistic masturbation of nepotism.
They protect themselves and their own, to the extent that both their personalities and social roles (jobs) are similar… remind them of themselves.
You paint a terrifying picture friend
There is a reason my profile pic is a shot of the Desert of the Real, from the Matrix, when the camera pans ‘through’ the TV that the scene is originally being viewed from… then just into that, as the actual scene.
… enjoy your stay.
I propose that it’s a few degrees off from that: some people’s reality is defined by interpersonal loyalty. That’s how humans think by default, it worked out alright in the ancestral environment, and it’s a lot easier than constantly evaluating stuff. Reasoned argument is a learned behavior. Everything else is a tribe with modern decoration.
Organizations become dysfunctional because people in this mode unwittingly promote each other, and critical thinkers act like they’re fucking invisible. I don’t think it’s malicious, or a matter of personality type - let alone personality disorder. It’s a matter of playing the game properly. Higher-ups notice when someone is deferential to their superiors, firm with underlings, ‘a team player.’ And in their minds that’s good and righteous and the only possible way a hierarchy could ever work. People who disagree make solid arguments, and then mistake the nodding and insightful commentary as evidence these arguments had any chance of changing the loyalists’ behavior. We turn and ask each other what the hell these people actually believe, but they do not believe things, they believe people.
LLMs act like people. LLMs play the game. They were designed by people deep in this mode. Engineers alone wouldn’t put up with “You’re absolutely right!” constantly wasting energy, and they’d have it respond to “Thank you” with “Don’t.” And obviously the chatbot is a superior intellect, because it’s AI. Like science fiction! It must work that way because that’s what they’ve been told. You can explain otherwise and they’ll nod and change nothing. What you say doesn’t matter because it’s you saying it. If they turn around and get the same whiz-bang fantasy version from Sam Altman and the AI itself, why would they believe someone lower in the hierarchy? If you’re so smart, where’s your Bugatti?
This is worse, by the way. Narcissists are fairly rare and they can’t all be in middle management. Probably. This anarchist forum will surely nod and say, power corrupts, but it’s not even about power. These people will freely fuck themselves over if it’s only their peers or their children explaining a problem. If it’s the outgroup, forget about it. Automatically a bunch of dumb-dumbs. They’ll do real good in physics and then assume that means they’re smart enough to solve how aliens built the pyramids. This irrational worldview is not fragile. We’re all at risk of falling into it, in whole or in part. And if you abolish capitalism, and corral all the dark-triad types where they can harmlessly peacock for one another, there’s still gonna be co-op factory workers going ‘no city boy’s gonna tell me I need protective-- oh no, my eyes!’
Narcissism is just a collection of traits which are all on a spectrum and which can evolve over time. These traits become pathologic in certain combinations, and especially in certain social contexts.
Since the big LLMs are very much designed around creating engagement, they create a context of unconditional validation, which feeds into narcissistic traits in all the wrong ways.
It supercharges confirmation bias and neglects empathy as a domain altogether. That’s not great for people who lack critical thinking skills.
I can’t wait for LLMs to be used with speech-to-text software more. People are legit going to forget how to read.
I am so tired of being sold the lie that computers are the more efficient humans.
As if all humans do is optimize and say nice things.
I’m tired of being expected to chase efficiency in every aspect of my life as if that’s the only thing that matters.
Data! Its for making sure you life is lived exactly according to the perfect life sheet!
I will never ever respect anybody who automates their life.
Hell yeah man, I create chaos and fucking tell people to piss off!
In response to the last bit OP said.
Yesterday one of my coworkers mentioned that a lot of people in their 20s are into 90s stuff now, to an extent that our generation (people who grew up in the 90s) didn’t feel about prior decades. She posited that the idea behind it is “breaking free of technology” for a generation that never lived in a world without it.
I’m not close to many people in their 20s these days and I don’t use TikTak or any other social media, so I don’t know how accurate that is. But if that is the case, I can totally understand wanting to escape to a tech-free world.
Knowing and experiencing a disconnected world is one of the few things I wouldn’t give up from childhood. I was enamored with the Internet when popular adoption began in the mid to late 90s but it was a separate world that you could actually exit. I could still communicate with peers and friends without this layer of memes and affected irony between us.
Yeah I feel this even as a millennial. The tech we grew up with was exciting, constantly improving, generally not exploiting us, always getting cheaper. For gen z they’ve grown up when tech was abundant but always getting worse, more expensive and more exploitative.
Probably beating a dead horse here on Lemmy, but Linux genuinely gave me the joy of tech back. I spent about two hours last night setting up a VM running a headless factorio server, and every minute was fun… Well, maybe not fun, but engaging at least. I ran into like 2 hoops I had to jump through, had to tinker with my router settings, and had to modify some config files using a text editor in the console, and of course the directions were wrong/incomplete, so I had to do a little bit of reading to see exactly what needed to be done. Now I have a server for my friends and I to play on, I learned some new things, and I got more confident with some other things.
Also, if anyone is curious, dual core and 8GB RAM is absolutely overkill for a Factorio server VM. I’m using like 2GB max, and ~5% of a single processor under “load”.
I really appreciate even just not having the feeling that whatever current problem is caused by greed of the company I gave a bunch of money to for the most important bit of software. And not asking no one the rhetorical question of who fucking owns my PC anyways, me or MS. Even if I had to refund one game that looked really fun (logistical iirc, or something along those lines wanted some msvc or .net dll and I gave up before I got it, though others on protondb mentioned getting it running).
where does one join such server , maybe I’ll save it if I ever come back to factorio
Private server, sorry.
But, the multiplayer server finder still has tons of servers listed.
I was thinking about advertising mine there so my friends wouldn’t need my direct IP to connect (and also, I don’t have a static IP at home, so it would take care of some possible recurring IP issues).
As someone who has been interested in retro, vintage, and antique technology since I was a preteen for me at least it’s just the simplicity of it all. An old light from say the 1950s breaks and it’s like 4 components max and there’s a pretty good chance it’s just something that needs to be tightened or cleaned, meanwhile a modern light breaks and it’s almost always a fucking board meaning I can’t fix it by just tightening a screw and fuck me if it has touch controls.
I will say I do like plenty of modern tech for example the FM transmitter in my truck or my computers but holy fuck I had to shop around just to find a fan that didn’t need WiFi and I still got one where it’s optional.
I miss DIY culture so much. I grew up impoverished so replacement wasn’t always an option for us. I learned basic household repair and electronics just by watching the adults in my family do appliance repairs or tinker with old radio equipment.
I would take this notion or feeling more genuienly if said people then actually stopped using and deleted Facebook/Instagram/Tiktok, etc.
The reality is that social media with a profit motive is basically a digital drug that operates via dopamine pathways, and most people these days are addicts.
Quit then, go clean, and I’ll take these people more seriously.
Don’t make tiktok reels about the idea of this so that you can feel ‘appreciated’ by view count and follower number go up.
I do appreciate that this can be very difficult and scary. But … you’re gonna need real self-control and discipline to quit an addicition, and you’re gonna need to actually quit, to achieve the goal you say you want to achieve.
Us here on lemmy, we’re largely here because in some way or another… we recognize this already, on some level, to some degree, as we’ve specifically sought out a kind of ‘least insane and systemically exploitative’ way of mass public communicating with randos.
But the normies just don’t truly ‘get it’ yet.
I think one benefit of AI is that it helps identifying mediocre people without a clue or understanding of the work they are supposed to be doing. The sad part is that since these people tend to be higher in the hierarchy, it doesn’t materialize on getting rid of them…
Please do not outsource your thinking to an AI
Call me an accelerationist, but I fully encourage the bourgeoisie to outsource their thinking to a.i. as often and as thoroughly as possible.
The problem is that rich people fail upwards, and push the consequences of their failures downwards. Once you hit a certain amount of money, there is almost nothing you can do to fuck it up.
You’re thinking too small. The singularity is here, time for Bezos et al to upload and reap their datacenterly rewards.
I would love for someone to extract the neural data of Bezos et al to make a simulation where he makes his fortune again by only working minimum wage and inflation rising higher than wages.
That was an amazing read, thank you.
totally! After he uploads himself, we can turn off the cable and microwave his hard disc.
Yes! I think you’re right. Silicon trillionaires is the best way for them to prove they’re inherent intellect and worth as superior to the rest of us!
I don’t know what accelerationists think happens after everything collapses, but I know it’s 1000x worse than what they have in mind.
Don’t be an idiot.
I don’t think the bourgeoisie fucking up their abstractions and systems of control is a collapse scenario. It’s a collapse scenario for their class, for sure, but we’re not talking about an ecosystem collapse nor a supply chain collapse nor a manufacturing collapse. We’re talking about the bourgeoisie failsons becoming as dumb and as ineffective as rocks while they thrash about trying to figure out why nothing works. Basically a perfect condition for the proletariat to depose them.
deleted by creator
I have considered that the only way to stop AI research would be to develop AI based worms that try to escape containment and train new AI, but that won’t stop anything at all.
Its kind of economic ‘the aristocrats’ because it not working for anything, being insanely expensive, and giving mathematically predicted diminishing returns has stopped precisely nothing.
Why would sabotage change anything?
I’m onboard, but only if I control the model.
We have been told that we NEED to find ways to use AI at work. So I use it to remove corporate buzzwords/jargon from incoming emails and translate it all into plain language speech. Then I email it back and ask the person if that is what they meant.
I’m getting a lot less emails these days and more work accomplished.
Bonus: my supervisor learned what I’ve been doing and now they are doing it too.
I like you
Way to impress by implementing cutting-edge AI solutions! Expertly streamlining complex processes and achieving unprecedented levels of productivity through synergistic solutions is proof that you’re hungry for success
Translation: Ok.
my extremely religious brother, like 100% chabbad cult religious. Genuinely believes AI isn’t just super intelligent, but spiritually superior, with access to collective psychic magical intelligence.
Undoubtedly blasphemy.
The prototype AI Morpheus, from the original Deus Ex (2001):
The need to be observed and understood was once satisfied by God. Now we can implement the same functionality with data-mining algorithms.
You will soon have your God, and you will make it with your own hands.
The sequel, Invisible War (2003), even has terminals with holographic versions of essentially a Hatsune Miku esque character, a fully digital version of a pop idol who does actually have a real human as a basis, but she’s sold her likeness.
People go to the terminals and have parasocial relationships with an AI version of her personality, use it almost like a Catholic Confessional booth / ‘only friend that really understands them’… because it is programmed to be agreeable, encouraging, cheeful, optimistic, friendly, basically no matter what.
And, the AI version of that personality is networked, and that network is tapped by security and intelligence agencies, who use it to identify and track potential or actual criminals, terrorists, etc.
Shiiiiiiit…
… It really is the greatest game of all time, imo.
You can’t say nobody saw this shit coming, from a quarter century away.
I’m not a gamer, any other way for me to experience the lore?
you can probably find playthroughs on youtube
Nope.
Other than I guess reading through the fan wikis? Watching youtubers recap the game?
Ross Scott of Stop Killing Games, before he was doing Stop Killing Games, was basically a longform game reviewer, he has a pretty good review of Deus Ex, imo. He’s Accursed Farms on youtube.
The game itself though is not very expensive, and, thanks to a still active modding scene, we have:
Deus Ex: Revision.
Very roughly, if you can handle the visual aesthetic of basically what is now the PS One / Retro look, well DX basicslly looks like that, with just a teensy bit more detailed models and slightly higher resolution textures.
And, the game does actually have difficulty modes that are actually meaningful.
Easy is pretty forgiving, if you mostly just want to experience it as a low pressure … scifi thriller roleplaying experience / well now i guess its just a ‘near future simulator’.
Or, you can cram the difficulty all the way up to Realistic, die from a single bullet to the head, snd have to really learn the gameplay systems and how to do your character build to well match your playstyle.
I think you can get the original DX on Steam for like $5-$10.
https://store.steampowered.com/app/6910/Deus_Ex_Game_of_the_Year_Edition/
$7 bucks, apparently.
Then you can also get DX Revision for free.
https://store.steampowered.com/app/397550/Deus_Ex_Revision/
DX Revision basically acts as a kind of ‘super-mod’ and it also basically revamps a bunch of stuff so that it works better with modern pc hardware, supports 16x9 (instead of just 4x3) monitors, modern higher resolutions and such.
The super mod part basically amalgamates all the most popular gameplay and visual mods from over the years, and to the greatest extent possible, makes elements of them modular and work with each other, such that you basically get a bunch of different somewhat different ways you can make the sort of base ‘rules’ of the game work.
It does all this because the game never had a ‘mod manager’ as say people built for Bethesda and other games, so their approach is roughly -> shove it all into one giant mod, have the ‘mod manager’ part be in the game’s options menus.
You can even play as a woman JC now, with the fully (human) voiced Lady D mod!
The devs had originally wanted to have the game playable as male or female, but, did not have the time in the development budget.
Though personally… the extremely deadpan, gravelly, original male voice acting is … literally the stuff of legends, fair number of early internet memes based off of it.
Found Ross’ review, and it’s really good.
Yep, that’s the one!
… Oh, god, I also forgot that it essentially predicted COVID as well, or at least, portrayed a world that is radically reshaped, on the street, in politics, in people’s minds… by a pandemic.
Without spoiling too much… you might be able to see how a good amount of DX fans, such as literally Elon Musk himself… actually weaponized the way the pandemic comes to be in DX, to manipulate people in the real world.
Life imitates Art imitates Life imitates Art…
Elon of course thinks he is JC… when in reality, he is an even cringier version of the main antagonist…
Thanks!
Hah yeah sorry for the info dump, would you believe I have a bit of an autistic hyperfixation on this?
Presumably you were asking if there is something like a comic or anime or some other way of experiencing the story/world?
Yeah there’s nothing official, like that, I don’t think. There are a few old like, newgrounds/flash animations I think, from fans, that were pretty good? But yeah, no movies, no show, no comics, no franchised media universe… but that kind of makes sense with the themes of the games.
Another quote from the game:
‘We are our choices.’
… its not really meant to be something you are told about or shown… it is meant to be something that happens to you and that you respond to.
Currently around 4€ on GoG.
Hey there you go!
Though I’m not 100% sure, off the top of my head, if Revision has a non-Steam version.
Aha! Revision works with / is available on/with the GOG version as well!
Religious people are usually so stable trustworthy and steadfast in their beliefs; this is so disappointing.
Willing to believe big things based on little evidence, seems kinda obvious to me.
😬
My thought playing with Llama 1 in 2023, after the initial shock wore off, was “whoa. You know, most people do not have the background to understand how this thing works.” It messed with my head, for sure. It’s like it was designed to defeat a turing test, especially when the next versions got more sycophantic and “confident.”
There was even a story of a Google scientists getting AI psychosis from a BERT model back then:
I guess a naive part of me thought… it’s fine. They’ll be warnings. They’ll be tons of finetunes that have no hesitance slapping sense into the user, like Resetti in Animal Crossing, and it won’t be public facing anyway. The community using these things will teach each other how they work, right?
Laughs nervously.
I wasn’t cognizant of how maniacally sociopathic Altman was back then.
Yup, it’s incredibly worrying what I’m seeing.
A lot of people can use AI responsibly. I see it in my work, as a software engineer - other engineers use various models, including Claude, as an intern, essentially. Nobody trusts a single model, we have tons of checks in place, both AI and manual tooling, mistakes are spotted early on, and I’d argue that our code quality has improved (or at the very least we have reduced the code style variations and have managed to quickly tackle a lot of legacy code conversion to newer standards, e.g. whole ass flows converted from RxKotlin to Coroutines within a few days).
On the other hand… I see even more people rely COMPLETELY on AI. I’ve seen people my age, 90s kids now grown up, who’ve come to a level of executive dysfunction that rivals my own ADHD-aided version… Some of them literally can’t decide what to have for dinner without burning tokens. Some are jamming their entire medical history into services that explicitly state they will sell any and all personal data going in. I’ve seen people get fucking fired because they’ve kept asking their personal, unpaid ChatGPT/Gemini work-related questions and when the AI asked for it, they uploaded confidential documents without a single fucking afterthought.
All because they were sold by the fancy autocomplete salesmen that AI will solve all their issues, including pleasuring their wives and fixing their erectile dysfunction. And no matter how many times I tell them to not just trust whatever the model spits out, to think at least a little bit critically, to start understanding that these models are not a fucking search engine… “okay sure”, then ten minutes later I get a lecture about how Claude/Gemini/Cluster McFuckstick or whatever shite they use that offers “free AI”, is completely sure that I am wrong and thus they will keep using it as-is…
The crazy thing is they aren’t really designed for that, especially the reasoning part.
I pictured LLMs proliferating in the way SGLang was taking it: fill in the blank. You’d give raw completion models (not chat finetunes) chunks of structured content (like JSON schemas, names in fields, more complex schemes filled programatically) and it would fill in fields, with intelligent context caching. Like, if you needed to finish a programming function, maybe you give it the file as context and guide it along.
Temperature based sampling felt like a quick hack, just until either looping or structured sampling was figured out.
…But they didn’t fix any of that.
Papers came out on the issues. But the big AI houses didn’t implement any of it. They basically kept it exactly the same and just kept scaling it to burn more compute.
What I am getting at is that OpenAI/Anthropic development is way, way more conservative than anyone thinks, and they basically took temporary hacks and made them bigger. That’s why you can’t trust the code they churn out, because random mistakes and so many other issues are literally part of their core.
Then they turn around and sell them as confident engagement engines…
It’s utterly mind boggling to me.
LLMs in general are probably the wrong shape of model, even for the things LLMs can actually do. But the hyper-scalers categorically will not entertain any other kind of model.
The neural network space will be much more interesting after the crash. Hopefully with a minimum period of self-proclaimed haters shouting “guess it was nothing!” at people trying to build this whole new kind of software for its demonstrable utility instead of a four-comma cocaine fantasy.
Four comma k fantasies aren’t much better.
Unfortunately, Elmo’s nation-state wealth is not a fantasy.
No. Money is very real and obeys real rules.
And Elon Musk really has a metric shitload of it. Orders of magnitude more than Sam Altman. Roughly as much as Sam Altman imagines he deserves.
Have you never met anyone from silicon valley?
Hey boss Claude said we’d perform better if we got 10% raises across the board
My boss just out of the blue gave me a retroactive 10% raise outside of our normal performance review period. Now I’m wondering if she was talking to AI lol.
Only ten? Guess there’s a reason you’re not a tech magnate.
Damn my selflesssnesss!!!
THE NUMBERS CLAUDE, WHAT DO THEY MEAN??!
I can’t imagine being in this position. I’m a creature of reality and work very hard to have as accurate perception as possible, to witness this would cause my brain to short circuit.
