When you think a magic technology is going to change 600 years of the business cycle, and you throw every dollar of capex into that hope, you get your railroad/internet/AI bubbleFor the midwits: Bubble /= bad technology, the spend ALWAYS overshoots the real demand. I’m sure AI will be revolutionary in 10 years
>>62179357So what happened in 1991?
If you don't know by now that a bad economy is good for stocks you really don't know anything about markets
>>62180030endless war in the middle east, i think?
>>62179357it looks ilke we decoupled. new paradigm inc
>>62180030small recession
>>62180033dubs of trooth
If AI doesn't pan out, America is literally out of the game. The crash will take the US out. I am wheeling about into shitty European stocks for like a year now. Your ETF's will get demolished too. If it DOES work then the US is going to rule the planet forever and we'll live like kings. Your money is kinda worthless in both scenarios.
>>62180566Even if it works out, it's possible that big AI companies don't benefit too much from itMy company is already looking into self-hosting our own llms after Claude started enshitification
>>62180566curious to see how the american chip fabs work out
>>62179357tell me what the graph is supposed to mean, fagi'm not seeing the red line diverge from the black line during the dotcom bubble or any other point, idk what it's supposed to mean
>>62180566there's no such thing as AI, LLMs are a parlor trickthey are ultimately limited by datasetsthe most usable ones depend on data from 30+ years of a cooperative internet, it's a miracle that happened and was recorded, there's no new data to use reallyneural networks are prone to overfitting if you make either the training dataset or the model itself (number of layers & neurons) too big, which is the main thing more GPUs supportsi.e if you make them bigger and bigger, they become useless after a certain pointbarring some actual breakthrough in the underlying tech (none has happened in like 8 years), which would essentially be a breakthrough in abstract math, they aren't going to get better from 1 trillion gorillion 10x hardware infrastructure 1 gorillion GPUs
that being said, idk how anything is gonna play out, i don't have a good enough understanding of the economic factors and retarded jew niggeryall i know is the most amazing things these consumer LLM companies do isn't in their tech, it's in turning (You) and everyone else and government officials into brand ambassadors & convincing everyone there's an arms race where none exists
>>62180634It's still really useful to increase developers' productivity at the very least
>>62180634While it's true AI is running into a bit of a wall when it comes to training data, it should be mentioned that it's possible to create large amounts of new data, especially for AI of other types than LLMs, and for this, computational hardware can help by creating synthetic data from computer simulations
AI isnt a bubble>>62180595claude code is too good i dont think you can replace that fast in your own small company.
>>62181244There are multiple good open source Chinese llmsThe idea is not to get rid of Claude from the start, but have other llms for simpler tasks
the line is going up, no?
>>62179357AI is once in a lifetime revolutionary productivity upgrade.While rest of the economy may have hard time AI will carry us over this dip.After rest of the economy recovers the bull run will be insane.
