Ashtear

Mass implementation is a mistake, and I suspect implementation in consumer goods is where the bubble’s bursting will be the most devastating. Recent news tells us Dell has figured that out already, and I don’t think it will be long before society decides it can’t tolerate things like AI companions for young children.

Ultimately, I don’t think widespread implementation with any sort of value will be possible unless someone figures out how to make effective prompt creation so easy anyone can do it. Everyone seems to think AI is just a box you press a button on and it’ll spit something out, but getting valuable output isn’t like that. Good prompt engineering and tool selection is hard, and it’ll have to be a trained skill for people working with the systems generative AI does stick around on.

The really unfortunate thing is LLMs are the perfect snake oil for sociopathic executives. They can provide something approximating meaningful human interaction to these lonely, workaholic MBAs, and from there, it wasn’t a hard sell to make them believe they could replace their pesky labor force, too. When you’re that far outside the real world, sycophantic illusions are seductive.

Having to use something like Windows IME on a phone for Japanese is nightmare fuel