Well, let's separate between the marketing fluff and the actual value. Currently it's doing with AI what everybody else does: using more data to make more detailed predictions about numbered data (in this case money). That in itself is already great.

If they achieve to recognize the car in a given picture, and maybe even the type of car, even though it's damaged, that will be really great. With that they may be better than the current best solution I've seen (FB's object recognition, which may not even be on as high a level as they present to us in a blog post). But estimating the damage, means recognizing something that is not a pattern. I doubt even classifying this into three categories like `~100bucks`, `~2000bucks`, `buynewone` would already be hard. I doubt that anybody could do it.

And here's the point: Yes, Chinese IT companies are picking up speed. And I wish them all the success. But they have yet to show something that is really innovated by them. This idea is great, but they need to prove that they can do it before I will clap for their achievements.