I find it hopeful see people trying to mitigate these risks while they have only just begun to be realized. There has been more than enough historical cases alr

I find it hopeful see people trying to mitigate these risks while they have only just begun to be realized. There has been more than enough historical cases already of inventions turning into something the inventor would dearly wish to have undone. One major current concern of mine, that this letter does not address, is AI for surveillance and social control. What is already being done in that regard is arguably military intelligence technology directed indiscriminately at entire populations, but the added element of powerful AI spidering over the data streams that go into places like the NSA Bluffdale facility is quite appalling. I think this is even harder to inspect for than autonomous killing systems, and even more difficult to avoid the development of, since much of the capabilities needed will likely be similar to what academia and data-intensive commercial sector will want to serve their needs. But the potential damage through empowering totalitarian control could well be comparable to or greater than an "AI arms race". I really hope the field will deal with this aspect too.