WHEN WILL THE FUCKING AI BUBBLE POP?
your hair will turn gray and fall out waiting for the market to become rational
>>106768423>WHEN WILL THE FUCKING AI BUBBLE POP?You are stupid if you think of it as a bubble.
>>106768432it is a bubble and it will burst, like your ass
>>106768423>>106768432the economy will de facto collapse while the stock market is glowing green. its no longer based in reality.
>>106768423Didn't you see Sora 2? You're here forever little myg0t nigga.
>>106768423Housing bubble needs to pop first. Wait your turn.
>>106768437>it is a bubbleIt's not, the bubble idea doesn't work, cloud is a bubble and it didn't burst.No, it's just the typical silicon valley scheme, hype up and then give a MVP, that will actually be used. They have done this time and time again and you keep falling for it and they keep getting away with it.
When dollar and the consumer can't take inflation anymore and they have to stop printing to sustain the markets.This has some ways to go.
>>106768470SV used to have 20 products and maybe 1 or 2 would become a behemoth and that was enough to keep them happy. Now they have 1 product in 20 different flavors and it WILL succeed and you WILL like it. It's clearly different from 15 years ago
Why do you folk hate AI so much?I've learned sooooo.... much using AI. Today, I realized I can conceptually put together an entire optimizing compiler, only thanks to AI.Then I went to Opus 4.1 and generated myself a 26-chapter book on design and implementation of an optimizing ISO C compiler in OCaml.I fed it loads of books and papers, and got myself a book, all on my own.This is fucking magic. And I am so thankful. Just 3 years ago, this was a pipedream.Be happy that we've made it this far as humans.
>>106768423when people stop funding it and it's required to turn a profit on its own, then all that perceived value will disappear because it was never there to begin with
I'm going to generate myself a 22-chapter book on type theory. I already got loadsa books on type theory, but I can't be bothered to read them.I just hope I can learn all the knowledge I desire before AI goes kaput.I need to learn database design and implementation as well.I dropped out after 3 and 2 semesters of SWE, twice, respectively. AI has been like a university for me.
>>106768508>when people stop funding it and it's required to turn a profit on its ownI agree with this expect.>then all that perceived value will disappear because it was never there to begin withThis is just a giant cope and willful thinking.
>>106768498>>106768534You're learning lies because AI isn't reliably correct
>>106768544> perceived value> perceivedSo being able to have all the knowledge in the world at your fingertips is not 'real' value?
>>106768558Why are you replying to me?I disagree with the last part of the post.There is no perceived value it has a value to it's users it's just going to be hard to create profits with it,
>>106768557If you feed it books and papers on the subject, it will not hallucinate. I make sure to double-check whenever it says something super-suss.I have enough foundation on the subject to know when something it says is wrong.I don't condone 'vibe coding' and I think the only people who vibe-code are sorry excuses for humanity, most webdevs.In reality, the types of programs that I make using AI, cannot be 'vibe-coded'. Like, imagine vibe-coding a fucking compiler, interpreter or a database.I just want a basic framework to base my own work on. AI gives me that. Especially if you give it enough knowledge in form of papers and books.This motherfucker gets its own Markdown syntax wrong. It skips '```' terminators, for example.As some dude said once, "It's not about using AI, it's about using it RIGHT".
>>106768590Uh. Apologies. I've been awake for 25 hours.
Shit is already crashing, why would you want it to your money will become worthless, it's already going to shit and it hasn't even started yet
>>106768557>You're learning lies because AI isn't reliably correctYou clearly have no idea what you are talking about, I have cross checked everything AI told me and it's by majority 85% correct {I'm talking about tech}that's same ratio as some lectors on school, here is an example they still claim fedora is redhat based, which of course creates confusion and they seem to not get it either that it's outdated info, and this is way before the AI boom.
Here's an example of Opus 4.1, with its "thinking cap" on, fucking up its own Markdown syntax lol.I don't know how these so-called "le vibe coders" operate. I think most of them generate JavaScript code that runs on the browser (aka Mobile ECMA-262), so browsers like Firefox and Chrome run their code and they are none-the-wiser.Still, treat it as a "fuzzy-wuzzy" sweater that keeps parts of your body warm, but if it gets too drafty, you gotta put some more layers on.
>Nvidia funds OpenAI>OpenAI funds Oracle>Oracle funds NvdiaBehold, this one design holds up the entire global economy. Baggies hate this one little trick.
>>106768627>85 percent of our planes land just fine thank you
>>106768627I think people need to understand what "fuzzy" means here. it does not mean furfag shit. Fuzzy logic means, for example, x is 25% true and y is 75% true. LLMs are fuzzy models. They make mistakes. It's up to you to be vigilant and find them out.But they are extremely indispensible when it comes to summarizing books and papers. I have very bad eyes, and I cannot read books and papers easily. Ever since LLMs, I've been able to digest at least 10 books and papers per week.Opus 4.1, especially with Thinking mode, is one step above the rest.
>>106768659No one is using AI to generate mission-critical code. People use AI to learn about wiritng mission-critical code.People who use AI to 'le vibe-code' are just dumb webdevs who don't know shit about crap.AI is also a good source of inspiration. Or, if you're undecided what your next step should be.It gets basic things wrong. For example, Opus 4.1 used a non-existent Stanza for OCaml's Dune just half an hour ago. But I was vigilant, and caught it.
I used one of them (not sure which one it was) to generate pseudo-code for emulation of the NES. Caught down my dev time by hours.https://github.com/Chubek/mikroNES/tree/master/extra/sudo
btw, most 'mission-critical' code is basically simple-af FSMs. And in fact, LLMs could be a lot of help in giving you this code.You won't ask the LLM directly for the FSM for, say, some plane or factory robot shit. You'll ask it for Coq or Isabelle/HOL code. These are "mechanical verifiers". If they cannot verify the code, the FSM is faulty.These sorta devices are verified bottom-up. The hardware is written in VHDL and verified, and the RTOS is written then verified, etc.LLMs can be a lot of help in verifying them.
>>106768817>LLMs can be a lot of help in verifying them.i think you misspelled "MBD" (model based design)llms are fuckign worthless at thatjust pause and give it a fink:why the fuck would you be using a statistical model to solve a deterministic one?
>>106768834>a deterministic one?a deterministic *problem
>>106768834I dunno man. I guess, you could have it generate the Coq code and you can run it.
>>106768834Besides, are SAT solvers not themselves NP-complete? Correct me if I am wrong here but would that not make them non-deterministic?
>>106768452>the economy will de facto collapse while the stock market is glowing green.that's how inflaton worksmoney is more worthless?stonks go up
>>106768852it wont generate correct coq just as much as it wont generate correct codembd would be better for that because it comes with stricter, easier verifiable constraintsbut one could also just write a correct system in mbdor the integration of ai generated mbd code would be much easiercoq is a representation of maths in general(i cant find the correct words also im self taught so i have a nigger tier vocabulary, please be patient)but mbd constrains the possible ideaspaceat its simplest you have inputs, you have outputs and these have to fit toegether. thats itits like graphical programming, but at an abstract leveland coq just doesnt, it doesnt verify the correctness by itself, its a way to express maths and the steps to solve equations to verify correctness yourselflike i said, i lack vocabulary, if i cant express myself satisfactorily ill make an additional effort>SAT solversbut thats what a coq based intermediary representation would bein my understanding, MBD is basically graphical programming in an abstracted manner.an intermediary representation from which code and tests can be generatedthast the difference, i see what you meanyou think in the direction of reformulating the problem into an equation you could solvebut thats not what an mbd does- or would do in the context of aino, its role would be a set of constraints that allows to verify correctnessand an intermediary, higher level representation that would help to cut drastically on the cost in tokens (in relation with context size)
I think the companies need to start making their own money in late 2026 or early 2027, so we have a while to go yet.https://www.wheresyoured.at/ai-bubble-2027/>One VC (who you'll read about shortly) predicted it will take 6 quarters to run out of funding entirely based on the current rate of investment, putting us around February 2027 for things to have truly collapsed.
>>106768852>https://chatgpt.com/share/68dea9d4-68d4-800f-bc13-1f80d7570172i was looking for specific names bc i know theres a couple initiatives that are working on that ideabut others are still writing papers about it, kek'came up with that idea myself once i heard what mbd was
>>106768423No one can see a bubble, that's what makes it a bubble, RETARD
>>106768498It's still an overhyped speculative meme bubble even if it's legitimate tech
>>106768498>I fed it loads of books and papers, and got myself a book, all on my own.>I'm going to generate myself a 22-chapter book on type theory.Isnt the point of LLMs to skip having to read the book? Otherwise why not just read the original books in the first place
>>106768423in 2 weeks, don't worry
>>106768461Housing cannot pop, that's not what a bubble is.Housing is driven by boomers that won't sell and builders pushing for condos instead of houses as well as real estate agents being scumbags inflating the prices.It's a classic case of supply and demand.AI is a bubble because it's all based on investors' money not actual product or results. That's what a real bubble is.
>>106768423Two more years.
>>106770323>Housing is driven by boomers that won't sell and builders pushing for condos instead of houses as well as real estate agents being scumbags inflating the prices.And then AirBnB, for one, and foreign investory, i.e. Arabs, China and rich dictators from around the world.
>>106770401>foreign investory, i.e. Arabs, China and rich dictators from around the world.I.e. more demand, but same supply, so...yeah.Great, eh?
>>106768423WHEN WILL THEY FIX FREETUBE REEEEEEEEEEEEEEEEEEEEEEEEEEE
>just download sp0rkeh...heh
>>106768423When investors realize what a dumb pile of shit it all is.
>>106770401>condos instead of housesbut condos are cheaper
>>106768431Pretty much. The market is neither rational nor conscious. The real value of all commodities tends to zero as automation increases, meaning the value in fiat currency of commodities will become increasingly sensitive to politics and external disturbances, and eventually it will cease to serve as an effective means of distribution and commodity socialization.>>106768558Your definition of value is non-specific. It of course has material value in use, but this is qualitative value that fills a need and cannot be measured with respect to other values. Whatever quantitative price it has on the other hand will always approach zero as techniques for training and building it require less human effort.>>106768423Who knows lol? You'd have about the same amount of luck predicting the behavior of a scared wild animal.
>>106768423The whole point of investing in AI is mass surveillance so never.
>>106768437Depends on context. There will be winners and losers. Maybe OpenAI will eat itself and Anthropic will run out of runway and be acquired, but LLMs as we know them now are a permanent addition to the landscape, and are never ever going away.
Money and the stock market is fake and gay with tons of "loopholes" and market manipulation tactics that serve to make funny lines go up.
>>106768423It will pop once institutional investors realize they're not getting returns on their AI investments. It's going to take a few years.
