/biz/ - Business & Finance

>>58633160
Your argument is missing the context that AI has been an "active area of research" since the 1980's and before. It has produced more disappointments than successes. The entire death of Lisp and Lisp Machines was due to DARPA/DOD pulling funding on AI: https://en.wikipedia.org/wiki/AI_winter

The only major new revelation was using drastically more sophisticated computer power to shift all focus to a machine learning/LLM paradigm, because modern computers can crunch tons of data. All modern AI relies on essentially stochastic processing of input against a large volume of data, whether it's from language/NLP (ChatGPT) or images/video (Tesla driver aides for example).

These systems "reason" by running a statistical match against things they've already seen. They do not possess fundamental concepts inherit to human beings or animals that make logical decisions, nor do they perform self-analysis. For instance, I've never seen an overturned truck on a highway, but if I do I'm sure as fuck not going to drive my car into it. Telsa "autopilot" drove right into it due to the machine learning basically have no concept for it: https://www.youtube.com/watch?v=ZQA1cY6n7YM

If I see someone wearing a shirt with a stop sign on it, I'm not going to slam on my brakes. No luck for a robotaxi: https://www.motortrend.com/news/can-a-t-shirt-stop-a-waymo-driverless-taxi-vehicle/

This is also why you can turn ChatGPT into a psychotic parrot. It does not have the ability to look at what it's saying and go, "hey that makes no fucking sense". It's just processing your input and running a match.

The modern AI hype cycle is mostly about securing venture capital funding for a novel technology by making it sound like the Next Big Thing so investors don't want to "miss out". They're relying on the dishonest idea that AI in real life is like AI in movies (a human-like intelligence), which, to paraphrase a book, is like inventing an airplane by trying to make a giant mechanical bird.