On one hand: this is a nascent field where there's lots of low hangnig fruit for picking, and enough uncertainty/hype to trick people into getting easy money.

On one hand: this is a nascent field where there's lots of low hangnig fruit for picking, and enough uncertainty/hype to trick people into getting easy money. On the other hand: AI startups don't make a lot of sense. The research and large data are the tricky parts. It can work; Google fumbled the ball and Midjourney as a startup managed to create a really good model. But then what? once the theory comes out, everyone else with a bunch of money swoops in and creates derivatives. It's great for consumers, but terrible for startups that are expected to shoot to the top and become monopolies like Facebook. It's also not a development problem where you can incrementally make improvements. An app or SaaS, you have features you implement them one by one to make it better. With AI, you could spend years researching RNNs for NLP only to have something like transformers come along and crush your dreams and make years of effort obsolete.