BrianV

I realize this is an odd topic in which to make a first post, but being as my day job is in the AI industry... The challenge with AI is 'the last 20%'. That is, writing an AI that can handle 80% of situations well is not that hard (well, relatively speaking). However, it's that last 20% of oddities and edge cases that the real world throws at you that become exponentially more difficult to train an AI to handle well. For generative AI (ie, ChatGPT or other AI that 'creates things'), the answer is simple - you won't see some of the complexities and subtleties that humans may put in their writing / art etc. This is why the Hollywood writers are pushing to exclude generative AI from the scriptwriting process, because they see a future where AI generates the first draft and they add the 'human touch'. For other AI, such as autonomous vehicles, it's a different story. Much of the research being done right now in autonomous vehicles is handling the myriad of different situations that can happen on the road. For example, different types of construction barriers, bad road markings, the unpredictability of other vehicles, animals on the road, and now what happens when all these combine. Progress will continue to be made in all these things, but at increasing expense and time.