Eric Siu says generative AI is overhyped now but underestimated long-term. Current tools like DALL-E can't edit images precisely; they remake them entirely. He predicts a two-stage workflow where AI generates first, then a separate tool lets you refine it, making the process actually useful.
AI is overhyped in the short term. It's underhyped in the long term. You know what the problem is with generative AI? It's like when you ask DALL-E or Midjourney to make something, you actually can't edit it. So if you say, hey, they make something for you, it's like, hey, can you change this for me? Or I like this, but change the shape of this over. It actually can't do it. It actually makes a brand new one for you. So basically what's going to happen, I believe, in the future is that you're going to have a situation where it'll create the initial one and then it might push over to like a model, like an actual engine where you can edit things, right? So it might create the initial like batch for you or the initial like inspiration, which will save you a lot of money, the concept, right? And then you can build on top of it. Right now you just can't do that, which is why it's not that useful at the moment.
