[Aperocky]

There seem to be non-stop AI hype since 10 years ago, but it does not seem to be a field that has high barrier of entry.

This time it does seem quite useful, but I think in a few years it will be commoditized and offered from all major cloud vendors, probably even get to choose the models.

[beebmam]

> but it does not seem to be a field that has high barrier of entry

Yeah for inferencing, sure. Try designing and training the models. It takes whole organizations to do that effectively.

[Aperocky]

Yes, but the most significant barrier is with the first, not for the followers.

I think it's a Tesla situation. Tesla did very well, and I believe OpenAI will be too, but everyone else will catch up and then some.

Once the first model rolls out, knowledge from that model would eventually disseminate. And it doesn't even need to be on a human level, industry and academia experts could probably make very good guesses on what's going on behind the curtains - and follow up. And we already have plenty of good models on our hands, some even open sourced.