The AI crash won't be like the 2008 crash, which affected everyone. It will be more like the dot-com crash, which mainly affected investors and not regular people. Basically, if you put a lot of money into the Mag 7 at its peak, you're going to be out of commission for a while. It could be 20 years before you break even. But if you're not a fomotard, you'll be fine.
k but what if the opposite happens
>>62179340I still don't know what AI actually does and what LLM is and how you can prove it
Ai is not a Bubble, maybe some semis but not mag7
>>62179397tesla is by far the most off kirk shit in history desu senpai
>>62179340Idk anything about stocks but this makes sense to me
Not happening. Too much is at stake. And we are building infrastructure, not domains.
if I had never listened to the sheer retarded fud from fucking morons like this I would have bought SNDK at 96 bucks when it was first shilled to me but was already up 300%FUCK these fucking NIGGERS
>>62179340These price spikes you're seeing are prices catching up with real inflation, not massive inaccurate valuations of these companies.
>>62179340>it could be 20 yearsI'll call your bluff. Time in the market beats timing the market.
>>62179340it is only the beginning, demand for compute is massive, new architectures are on the horizon and need large-scale training, every medium-sized business wants to use open models, every software developer is hooked and will never go back to writing code manually, and big boy model providers are just starting to raise pricesit could be 20 years before it stops growing and starts crashing
>>62179676YOU LISTENED TO FUDDERS AND SEETHERS YOU ABSOLUTELY RETARD YOU DON'T DESERVE TO MAKE IT
idk if this is going to happen or not but it's really starting to feel like a bubble. I liquidated everything and I'm sitting on cash. Now that I've done this the market is probably getting ready for the biggest bull run ever
>>62179787Explain why you think it's a bubble
>>62179810desu it's less about AI, and more about the global energy market. You should know I'm almost purely going off vibes, but I wanted to get out before any craziness
>>62179817Okay
>>62179810not op, and i don't think it's a bubble, but going off what's happening in this presidency we are just blatantly pumping and dumping at any opportunity. the rest of the world is starting to get uppity so i think he's gonna clap back and make the dollar stronger...also means eating shit because the rest of the world has to eat it harder. then we get a new fed and let the printer rip one more time and after that things need to settle down. the younger generation are increasingly gonna get cooked.
>>62179779Nothing in stock trading is about 'dserving' you moralistic twat.
If AI is a bubble it would actually impact everyone simply because of how much money is being poured into hardware that is pumping the GDP. With dotcom there was never any actual revenue behind the stock prices, hence the outlandish P/E ratios. People are actually spending money on actual goods and services this time around and all of that will crater if there is a bubble.
>>62179901>the younger generation are increasingly gonna get cooked.nobody cares about anybody younger than themselves
>>62179901>he will just make the dollar strongerThe massive money printing currently propping up stocks is already sliding DXY down this year. If they cut rates then USD will just nosedive
>>62179676Always believe in yourself fren
>>62179751>new architectures are on the horizonspill.because i've been waiting for something that isn't a dressed-up perceptron for over 30 years.
>dat AI aint useful (yes i use it daily) -/biz/
>>62180189you made a lot of money investing in chip stocks though right? since you understood this all so well and so early on?
>>62180189deepseek invented a way to make models more efficient https://arxiv.org/pdf/2601.07372but to use it they need to retrain models from scratchrobotics needs better and better world models and jepa is a promising new architecture for 3d space generation>waiting for something that isn't a dressed-up perceptron for over 30 yearswhy stop at perceptrons, midwit? go deeper in your cope and reduce it to dressed-up matrix multiplication
>>62180193The internet is useful. Didn't stop the dot-com crash from obliterating many portfolios.
>>62179340>the dot-com crash, which mainly affected investors and not regular peopleLMAO the recon king has arrivedall hail the retcon king
>>62179340I’ve exclusively bought shares in energy infrastructure and manufacturing, which gives me some exposure to AI, but with the Belt and Road Initiative and the Global Gateway project, I’m predicting a 50% return by EOY.
>>62180230>dressed-up matrix multiplicationcorrect.LLMs are decent (if somewhat unreliable) summarizers, and it's a good architecture for simple, repetitive tasks that don't require precision or reliability (which we've known about small "artificial neural nets" for decades, we just didn't have the hardware to handle language until fairly recently).i'm not suggesting they aren't useful tools - what i want is actual AGI. they won't get there.ultimately you can't get around information flow - an entirely feed-forward system cannot be self-aware (and thus cannot adjust its actions based on introspection), because information flows exclusively in one direction. it fails any definition of intelligence that includes self-awareness or cognitive executive function.we also know for a fact that backpropagation is not how neurons work (this is a simplistic idea from the 1950s, before much of modern neuroscience). it requires total top-down control of the network and the ability to shut it down to update weights, whereas neurons operate locally (unless you presume some sort of instantaneous quantum/psychic field between them... which i don't) and continuously, updating long-term memory while still active.we're gonna need a FUNDAMENTALLY different architecture to get AGI, but every time somebody says they have one it's some extra tacked-on thing to LLMs (like RLHF, or "reasoning" models that don't actually reason, they just embed an invisible summary into the context window to improve next token gen accuracy... sometimes).i crave actual new architecture, not calling the same architecture with a coat of paint "new architecture.">>62180195honestly, i overestimated the intelligence of silicon valley. i didn't think they'd get this delusional this quickly. i suppose i wasn't paying enough attention to them in the 90s.
>>62180230i will say, this is much more interesting than the usual "new architecture" declarations. it still has a lot of the same inherent unsolvable problems, and still goes down the feed-forward LLM direction, but it's at least novel.it's a new architecture for handling perceptrons/matrix multipliers, but when i say want new architecture i don't mean i want more complex wrappers. the core LLM architecture they all rely on is insufficient for the task of AGI, and it has been extremely frustrating watching basically zero research or funding go into anything else.
There is a bubble but it's just getting started. Just because it's all hype and little substance doesn't mean you should avoid the trade. Dance at the party but just make sure you leave before the music stops.
>>62180285>>62180301spam filter tongue my anus
the scaling "laws" have been wrong for nearly two years at this point. if scale alone was enough, no new wrapper architecture would ever be necessary.
>>62180329scale alone is enough, we ran out of data and ran into the wall with existing hardware and power grids, waiting for new hardware to drop, new datacenters and power plants to be builtresearch makes models more efficient or easier to train, but you could get the same gpt 5.5 level of intelligence with 2.5m tokens context from a huge dense transformer with no wrappers, the problem is that if you owned every existing GPU and TPU, a naiive pretraining run will last 8 yearssame is true for perceptrons or any other mechanism of approximating functions, without the attention mechanism (a genius invention that you foolishly call a wrapper) those will take 800 years to train on all existing hardware, if they will even converge on anything
>>62179340Ai bestest thing since sliced bread!
>>62179340Youre a mouthbreathing subhuman. You dont know what market capitalisation or automatical rebalancing means.
>>62179787Next bull runs got to be the last. Doubt the gains will even be massive anymore. Retail is sick from nfts(2021) and Solana(2025) and crypto seems to be detested by the normies even more than ever with the whole AI is draining our waters and data center hate. I bet only alt coins will rise and even then it won't be anything insane like 2021
>>62179901>clap back
>>62179340crucify tech CEOs.break the skulls of tech CEOs.burn tech CEOs to a crisp.etc.
>>62180285you overestimated your intelligence for sure lmao. better luck next time.
>>62180877LLMs have been around for years. they've proven useful in generating boilerplate, acting as glorified search engines or chat bots, and doing basic analysis of language -- but that's the extent of their usefulness. it's a probably weighted sentence constructor. if you can't conceptualize the limitations imposed by this, you're likely an unsophisticated retard who is readily fooled by any sort of parlor trick.
>>62180897i invest to make money. better luck next time to you too lmao.
>>62180877You can't do anything without asking a retarded hallucination box its take on something anyone with any capacity for thought could have either worked out on their own, or at least looked into and formed a semi-organic opinion on. "AI" users are the lemming bulk of any given general population, and now are only getting more numerous as this backslide becomes acceptable.
>>62180922and this helped your brokerage account go up? what is your point? is this /biz/ or peoples diaries?
AI is not a bubble. It's the core of the fourth industrial revolution, for better and for worse.
>ur so stupid>everyone stupid retard stupid>listen to my sad story and opinion on 4chan /biz/ has gone down hill lmao
>>62180285>LLMs are decent (if somewhat unreliable) summarizerI only had to read one sentence, to see how out of touch you are from the AI revolution. These things can already create full-on Apps, platforms and all sorts of front-end complex products. See Claude Code, for example.
you retards are going to lose so much money jerking each other off over useless stock investment choices, how anyone looks at the stock market today and thinks they can predict the future in this crazy ai, trump, crypto world is beyond me
>>62180933>sir do the considerationing of the tool called as CLAUDEhow many rupees do they pay you people? it's all over 4chan, reddit, linkedin, etc.
>>62179379>what AI actually doesIt produces slop>what LLM isThey're fancy word predictors using statistical propability. >>62179751>demand for compute is massiveOnly because a few AI companies hoarded GPUs.>need large-scale trainingWhich has already reached diminishing returns. >every software developer is hooked and will never go back to writing code manually,Leading to more code slop and bugs>big boy model providers are just starting to raise pricesWhich hardly anyone is going to buy, especially when there are free Chinese models available.>it could be 20 years before it stops growing and starts crashingMore like 1 year as OpenAI misses more revenue targets, and is nowhere near profitability.
>>62180285>we're gonna need a FUNDAMENTALLY different architecture to get AGINot going to happen this century, or even ever. Humans still don't know how consciousness works, how can software engineers be expected to invent artificial consciousness?
>>62179379https://www.youtube.com/watch?v=LPZh9BOjkQs
>>62179901If they move industry and production of things back to the US and want to be competitive exporting goods, a weaker dollar is desired.
>>62180273>energy infrastructure and manufacturing,What companies?
