Regarding the risks, it would seem that (1) we can mitigate the impact of automation through some kind of basic income and (2) we can prevent rogue, malicious, biased, or otherwise harmful AI through appropriate thoughtfulness into instilling ethics and avoiding biased datasets (see, for example, Nick Bostrom's book Superintelligence.

But I think the biggest risks are the unknown unknowns and the unclear timeline. I'm currently reading the book Founders at Work, interviewing startup founders in 2007. It was just 10 years ago, yet virtually none of them say anything about smartphones, and the section on Research in Motion talks about BlackBerry being the most ubiquitous mobile device in the world. This was just 10 years ago, and nobody was talking about one of the biggest trends in technology that was coming, mobile, with obvious signs if you look back. So what, regarding AI, will look obvious in the future, but isn't obvious today? Is it even possible to predict? What will the timelines actually be?

As for the intersection of blockchain and AI, I absolutely think this is a fascinating area, and brings us closer to truly decentralized autonomous organizations. For example, see the AI-blockchain hedge fund startup Numerai.