AI is a grift propped up by blind VC funding. The cost is not justifiable and there’s no evidence it will turn a profit at any point in the foreseeable future. It is a textbook case of mass psychosis. The bubble needs to pop and the technology industry will only continue to deteriorate until it does.
>>108437493Yeah but until then I get to abuse retarded LLMs for free while the companies keeping them up and running lose money due to my dumb queries, it's amazing
got any more scathing and fresh original takes for us?maybe something about water usage, haven't heard anything about that one beforeoh i thought of something! This is kinda like a... oh man im really working my brain here... it's kinda like that thing, you know... that, OH yeah it's like another dot com bubble! I need to contact the wall street journal
It's more like an overhyped but still useful technology.
>>108437596Useful doesn’t cut it. It’s too expensive.
How much subsidizing is actually happening for let's say a $200 a month pro plan?
>>108437651Only the big models with big context windows. You can run models on your own machine that are still useful. But yeah, they're not great as a business model.
>>10843766220.000$
>>108437747sama said their inference business is profitable, its only the training that pushes things into requiring additional investment
>>108437993>scama says
>>108437712They're currently not great as a business model, but you can absolutely run inference of capable models quite economical, especially if the centrally hosted models were focused less on serving subsidized frontier models.Once models become longer lasting, implementing them in hardware really makes them viable, here a proof of concept: https://taalas.com/products/
>>108437493It's turning a profit right now as companies pay huge subscription fees and give their employees token
>>108440429It is obvious that making ASIC hardware (if I am even using the term ASIC properly here) is the solution to making them viable tools, but the main issue is that eventually the best models still aren't anywhere near automating even a McDonald's drive through window order taker role. When viewed as a productivity booster instead of an eventual job automator, VCs lack the excitement and potential gains needed to keep funding improvements. It is also a big problem that every single scientific study on how much productivity the LLMs give to programmers shows that teams are actually slower using the tools than using their old approaches to coding, but workers almost thought that their productivity was actually improved. The world is rapidly reaching a point where sunk-cost fallacy won't keep money flowing into these projects much longer, and it seems like the technology will likely end up a dead end like Google glass, Metaverse, AR, etc. The main issue then becomes that we spent trillions getting to this dead end, which we have never had such irresponsible funding of emergent tech in history, so the economic crash as the bubble pops, likely being compounded by the energy costs skyrocketing because of Iran, will plunge us into a depression for some number of years greater than 4.
Does this look like a grift to you?
>>108437493AI advocates fall into seven categories:>pedophiles who use it to generate images of CP>AI executives trying to sell their scam>60-80 IQ non-AI executives who think it will automate all employees>spiteful NEET retards who think it will automate all employees>mulattoe teenagers who use it to cheat on dey home-werk>codetrannies who think AI's adeptness at creating boilerplate extends to all jobs>retards who confuse proven and useful technologies (machine learning/vision) with LLMs
