I did not say future AI is limited. I said our current approach is flawed and very unlikely to ever result in a true AI. Whenever we do build that AI, it won’t be with a better version of the tech were using now, but a very different approach will have been taken.
It’s the same way we realized you couldn’t build a true AI by just trying to create enough if/then statements. You can make some fancy software, but the approach was inherently flawed.