David Gerardi am disconcerted by how many techies who saw through crypto *immediately* have fallen for AI krokodil
gaytabase@davidgerard i wonder if there's any correlation to how much importance you place on fundamentals. if you value those, no amount of apparent good behaviour will convince you that something fundamentally flawed is good, but otherwise, as the illusion gets better you're subject to all the usual cognitive defects, particularly the ones apparently smart people feel they're too smart to fall for.
Cap Ybarra@davidgerard i can't explain it yet. maybe they saw a more immediate profit available in llm boosterism. maybe it attacks a different center of the brain.
Azuaron@cap_ybarra @davidgerard There's a wide swath of techies that succumb to the allure of "usefulness". Crypto is easy for them to see through, because it has no fundamental use case that justifies its expense.
LLMs, on the other hand, have many apparent uses. Not only that, but techies love using tech to solve non-tech problems--often because they don't understand those problems, so assume the solution can be easily solved with tech. LLMs present themselves as the "anything solution", and it can badly do so many things (but: it does do them) that many techies just point at all the things it can do and go, "See how useful it is!"
LLMs are Dunning-Kruger wet dreams.
And what I'm learning is, so, so, SO many professional programmers are actually shitty programmers that can't even tell the difference between error-ridden slop and maintainable code.
Alien software, human hardware@Azuaron @cap_ybarra @davidgerard "SO many professional programmers are actually shitty programmers that can't even tell the difference between error-ridden slop and maintainable code."
This also explains the existence of nodeJS and every js-frontend framework out there.
(Only half joking, i think the causality goes the other way: "everyone uses react, must be good programmming skill" -> everyone measures their skill on crap -> machines produce the crap -> machine output is great!
fancysandwiches@davidgerard well, the potential utility is way more obvious. Crypto is and always has been pretty worthless, and also involves a ton of risk. You can use LLMs without taking a big financial risk. You can spend a weekend with a free trial and build an app that you never release and costs you nothing. With crypto you have to put up money, then you do something like dump it into a defi project, then inevitably lose that money, or fall for a wallet scam, or whatever.
JP@davidgerard yeah i think the most common pitches for either appeal to different self-identifications and impulses. crypto was/is bag culture, everyone around you is a cash machine, entrepreneurship is the only remaining form of social mobility so best get hustling. the LLM hype that targets programmers hardest is about appealing to "intelligence", productivity/status insecurities, and becoming a manager. most of them were already comfortable so the threat is the removal of that comfort.
@davidgerard what they have in common is their ability to select for people with holes in their social-political consciousness who can be persuaded to dismiss arguments about social externalities.
@jplebreton also ~100% of crypto bros fell for the chatbot
Jer@davidgerard @jplebreton crypto bros are about the hustle. GenAI is the current hustle.
Same group of guys will move to whatever the next big thing is. If there is no next big thing when GenAI investment collapses, they'll be the guys either jumping out windows or consolidating their money depending on whether they were able to sell at a high or not
@davidgerard building a grift around a product that will tell you literally anything you want to hear is inevitably going to be way more broadly persuasive than crypto's "get rich doing nothing (actually you have to work a shitty and obviously evil telemarketing job)" pitch. so it's not too surprising that the crypto venn circle is mostly nesting inside the llm circle.
EntitasThe amount of damage done by techbros convincing themselves that they are very smart, while being the dumbest and most credible people alive for any sort of buzz that vaguely promises to tell them they're special, has been catastrophic for society as a whole.
Shephallmassive@TheEntity @jplebreton @davidgerard the west has a very bad history with tech bros and the far right - no one wants to look at the lessons learned but every important powerful committe org or board without women on it is like a dead canary in a coalmine. On a screen near you at the moment bros making decisions as the women, children and powerless are pulled from the rubble..
¡a!n@davidgerard the lack of instant gratification with crypto vs AI’s sycophantic instant dopamine hit
Tobias@davidgerard I think its because it has the potential to generate cat pictures (or any equivalence thereof) which crypto can't. It's something extremely concrete.
Jess👾@davidgerard Cryptocurrency thrives on greed. Vibe coding thrives on imposter syndrome.
Laurent Bercot@JessTheUnstill @davidgerard It's not imposter syndrome if... you know what, I'm not going to finish the thought
luca@davidgerard They've been easy fooled by proposing it as revolutionary change that you can't be behind. Driven by bribeing a handful of relevant influencers.
Scotty Trees@davidgerard it’s not really the same comparison though is it because crypto you had to invest in it you had to go out of your way to use it but with AI it’s kind of just sprinkled and everything so I don’t really see your point
Enola Knezevic@scottytrees @davidgerard "AI is sprinkled everywhere"
ChrysanthosI think it is partly to do with the brainrot from years of seeing "tech advance quickly" even the anti-AI people were early on mostly assuming that this tech was legit gonna be replacing artists etc.
I must admit I have myself become disabused of this notion now. We are long past the days of seeing video game graphics and mobile phones advance every year. Some technologies are just dead ends or cap out early on like fridges. Turns out LLMs are a dead end tech and the corpos are just horny to replace us with them.
@chrysanthos yeah, they really obviously hit the top of their s curve in 2023 and *everything* since then is logarithmically more effort for small linear gains
@davidgerard Even people like Cory Doctorow still have a serious case of the "tech libertarian brainrot". It's a real problem with the techhead world. You'd think with how much Star Trek we all watched growing up we would've learned that message about technology must not mature faster than morality.
Scott Michaud@chrysanthos @davidgerard I'm guessing some (not all) of the priming to replace workers is related to post-Covid remote work, revival of unions, etc. The seeds of spite were fresh in the memory.
Korgie D. 🏳️⚧️🏳️🌈@chrysanthos @davidgerard It has never made sense to me that some people think it could continue long-term or was sustainable. But even when tech was still advancing quickly, it caused many problems like a major increase in consumerism and transformed us into a throw-away society. It made so many people not appreciate what they had, and instead had them constantly needing the next thing. And advancing hardware made software get lazy and inefficient, where overcoming that with brute force (new hardware) just killed the environment with old e-waste and ever-increasing power usage of newer stuff.
Also, I'm not sure why anyone ever believed that there could be a desire for AI to replace the arts :/ or to replace anyone or anything, really.
J. R. DePriest 
:EA DATA. SF:It has niche use cases that are truly beneficial but that's not what's driving it.
If you want pervasive surveillance and autonomous systems where the inaccuracy is considered a benefit and a sledgehammer with which bludgeon your creative talent into submission or poverty, this is the tool for you!
Damn the environmental impact!
felix (grayscale) 🐺@davidgerard LLM agents are tapping into people's desire to have obedient minions who will RP a world that responds to personal control, without the messy complexities of human minions
Olivier Roux (moose)@davidgerard @cstross because crypto was tech. AI is “magic”.
Jonathan HendryMoney-motivation vs Laziness. Fewer of the former, more of the latter.
Peter N. M. Hansteen@davidgerard Both are about throwing insane amounts of hardware at crappy code, financed by rubes who understand neither?
Orava@davidgerard
I think, a "cultural" aspect plays into this. Before Bitcoin (and probably still, tbh) almost nobody had any idea what a "cryptocurrency" is even supposed to be.
AI otoh has been a significant part of very popular stories for decades. When people hear "AI" they have an instant idea what they imagine that to be. An idea that has been culturally reinforced so strongly, that the idea itself disguises the reality of what the so-called "AI" before them actually is and does.
I think, a "cultural" aspect plays into this. Before Bitcoin (and probably still, tbh) almost nobody had any idea what a "cryptocurrency" is even supposed to be.
AI otoh has been a significant part of very popular stories for decades. When people hear "AI" they have an instant idea what they imagine that to be. An idea that has been culturally reinforced so strongly, that the idea itself disguises the reality of what the so-called "AI" before them actually is and does.
PassengerI've been wondering about this. I think it's because machine learning (in the broad sense, including other technologies like DL) is fun. It's really technically interesting to implement ML algorithms to approach problems, and even more interesting to try to solve the problems those ML algorithms cause.
Contrast this with blockchains, which are pretty boring from a technical point of view.
A large part of why I write code for a living is that it's fun. Non-technical people might not understand this, but designing code (especially optimising code) is basically a puzzle video game.
So I can see why people with brains like mine might get swept into wanting to believe that the fun project to hack on is also going to be worldchanging.
catchI think this would be true if people were writing LLMs, I know one person who started writing his own to run locally. Not for me but I can see where that comes from.
Using LLMs as a service just does not fall into this for me though, about as far from profiling and performance optimisation (which I like and agree can be puzzle game like) as I can think of.
@catch56 @passenger ~100% of the guys saying "LOCAL MODELS LOCAL MODELS" are on the $200/mo Claude subscription
@davidgerard @passenger not sure you even need the ~ there. Even if one doesn't have it we know there are people running multiple subscriptions to cancel them out.
Yeah this is a fair point and I think beats my point.
Dźwiedziu@davidgerard
No “100% local «AI»” guy has shown me his model.
Landa 
F
d@nny disc@ mc²@passenger @davidgerard blockchains are indeed terrible which is so frustrating because the application domain they claim to cover is perhaps the most interesting thing in the world to me. when i have a better handle on the build and packaging shit i will be releasing more work in the area of anonymous networking
katrina@davidgerard
With "AI", there is the perception that it sort-of works at the moment and they just need to fix a few bugs and everything will be perfect. Trying to challenge that argument is what I find most difficult.
With "AI", there is the perception that it sort-of works at the moment and they just need to fix a few bugs and everything will be perfect. Trying to challenge that argument is what I find most difficult.
@katrinatransfem my usual answer is "you've been saying 'next model bro' for three years"
Michelle Bacon@davidgerard @katrinatransfem It's just like the "self-driving cars are X months away" pushers who won't admit they were wrong. The reality that not all tech is viable seems to escape them.
Vesipeto Vetehinen@michellebacon @katrinatransfem @davidgerard After many years of unwarranted optimism it does look like self-driving cars are now scaling up quite fast though. Waymo's already in business in 10 US cities with more TBA this year and China has theirs too.
RAK@davidgerard @vetehinen @katrinatransfem @michellebacon: Also, considering the terribleness of these even with human intervention, this is a failure of regulation more than a triumph of technology.
@raktheundead @vetehinen @katrinatransfem @michellebacon we've replace one driver with 1.8 drivers, a stupendous success in automation
@davidgerard @raktheundead @katrinatransfem @michellebacon What are you referring to with that number? I believe they said in the congressional hearing that they have about 70 remote operators on duty at a time. Those are operating 1000+ vehicles simultaneously on the road.
There's still many other things that need to be done when operating a taxi fleet that Waymo doesn't automate and need additional employees for currently of course. Charging is one thing that they eventually probably will just like the Chinese Apollo Go has already done, they have charging stations that automate swapping the robotaxi battery to a fully charged one.
There's still many other things that need to be done when operating a taxi fleet that Waymo doesn't automate and need additional employees for currently of course. Charging is one thing that they eventually probably will just like the Chinese Apollo Go has already done, they have charging stations that automate swapping the robotaxi battery to a fully charged one.
mx alex tax1a - 2020 (6)@davidgerard @katrinatransfem and the response we get to that is "noooooo! why can't you see the upward trend of capabilities the model is gaining???"
@katrinatransfem @davidgerard yes but eliza doesn't make these numpties feel productive, and doesn't tell them "you're absolutely right!", and anyway, the new model has a widened soundstage and brighter treble
@atax1a @katrinatransfem @davidgerard
Look how smart I am! See, I paste my horrible ideas into the head-patting machine, and it tells me I'm right and a genius and that the idea will totally work. You should try getting your head patted by the head-patting machine too.
europlus 
Eliza - Copilot 1966 style@europlus @katrinatransfem @atax1a @davidgerard
*** ELIZA ***
Original code by Weizenbaum, 1966
To stop Eliza, type 'BYE'
SAY, DO YOU HAVE ANY PSYCHOLOGICAL PROBLEMS?
~*~ the benchmarks ~*~
Elias Mårtenson@davidgerard @katrinatransfem there is something fishy about the most recent influencer campaign. They even got Knuth to mention a specific product by name.
It's like they did psychological profiling on individuals, and then give them exactly what they need to praise them, and not their competitor (and of course push the concept as a whole bit only secondarily).
Mae@loke @davidgerard @katrinatransfem it's odd how it feels like it all happened at once. I don't buy into it being a grand conspiracy, but I also wish there was an explanation
@Mae @davidgerard @katrinatransfem not a conspiracy, but a well executed marketing campaign, I think.
@loke @davidgerard @katrinatransfem I think an undisclosed guerrilla marketing campaign of this size qualifies as a conspiracy