I’ve never read an article about AI before this one. I feel qualified to comment because that places me squarely in your target audience for this piece. I found it to be an interesting introduction to the 3 questions you highlighted for discussion. I found the discussion of “friendly AI” particularly compelling/scary.
I do have a couple of comments on style that might be helpful:
In the opening section, the three paragraphs on what you are not planning to do may be useful signaling for people who are fed up with some elements of the AI discourse. But as a newcomer, it made me feel as though you were on the defensive from the beginning. I also found parts of these paragraphs a bit confusing, in contrast to the rest of the article which is clearly aimed at newcomers. E.g. I don’t know what a classical Von Neumann computing architecture is, or what it means that you aren’t assuming the continuation of Moore’s law.
Section 1 (“from here to AI”) is a good introduction to how to think about the tricky problem of ‘when is it coming?’. But it wasn’t clear to me how your conclusion that “there is a significant probability that AI will be created this century” follows from section 1. It might help if you were to spell out which of the seemingly difficult AI benchmarks have been reached (are these the Accelerators you discussed before?).
I’ve never read an article about AI before this one. I feel qualified to comment because that places me squarely in your target audience for this piece. I found it to be an interesting introduction to the 3 questions you highlighted for discussion. I found the discussion of “friendly AI” particularly compelling/scary.
I do have a couple of comments on style that might be helpful:
In the opening section, the three paragraphs on what you are not planning to do may be useful signaling for people who are fed up with some elements of the AI discourse. But as a newcomer, it made me feel as though you were on the defensive from the beginning. I also found parts of these paragraphs a bit confusing, in contrast to the rest of the article which is clearly aimed at newcomers. E.g. I don’t know what a classical Von Neumann computing architecture is, or what it means that you aren’t assuming the continuation of Moore’s law.
Section 1 (“from here to AI”) is a good introduction to how to think about the tricky problem of ‘when is it coming?’. But it wasn’t clear to me how your conclusion that “there is a significant probability that AI will be created this century” follows from section 1. It might help if you were to spell out which of the seemingly difficult AI benchmarks have been reached (are these the Accelerators you discussed before?).
I hope this is useful
Yes, thank you for your comments, and for noting that you’ve haven’t read articles on AI before!