Bottom line is: LLMs are undercooked. They are auto complete bots, but there are people out there that think they are like having a really smart friend.
More like having a friend who knows everything about everything - but when that friend is fact checked, they are mostly incorrect. Yet companies are investing billions of dollars into rollouts of AI features - without them being in a "finished" state, capable of interacting with humans.
It's similar with self-driving cars. Companies can't even manage those robots that patrol grocery stores to stock shelves. A little 2 foot robot has a fit when a human enters the aisle - because robots are not programmed (and researchers haven't found a good way to program them) to deal with autonomous decision making agents (Humans) that might do something other than what was expected. Actual researchers in the field of AI and robotic vision know that we are leagues away from what is promised - But there's a whole lot of hype around self-driving cars and AI tools, and a WHOLE lot of money being poured into achieving very little.
Part of me gets it: You own a company that promised self-driving cars. You have to believe your own bullshit to entice investors. You roll out a product to market despite it's obvious flaws because you need to recoup capital and sell some of your products to some rubes so your investors don't sue you into bankruptcy. I get it.
|