The AI community treats potential customers as invaders. If you report a problem, the entire thing turns on you trying to convince you that you're wrong, or that you reported a problem because you hate the technology.
It's pathetic. It looks like a viper's nest. Who would want to do business with such people?
Actual promising AI tech doesn't even get the center stage, it doesn't get a chance to do it.
As godawful as this brute force, "change the laws because China!", 24x7 assault of LLM hypelords has been (and I love GP's analogy about finding a helicopter useful), and its been preeeeety unpleasant, there is a silver lining.
The compute and tooling needed for all the other under-explored, nifty as hell, highly accessible ML/AI stuff is like free by comparison now: there are H100s floating around for around a dollar an hour, L40s for sometimes pennies on that dollar, and like, all of neural machine translation or WaveNet era speech to text or resnet style transfer was done on like, a thousand bucks in today's compute. Lambda Labs has a promo on GB200 where its cheaper than H100!
And there's " plenty of room at the bottom": Jetson boards and super cool autonomy stuff like that is Raspberry Pi accessible.
I'd rather they didn't feel the need to like, take over the government to get terrawatts of "just make it bigger", but given that's sort of already happened, I'm looking for what opportunities are created by such monomania.