Dilution of AI risk efforts
Over the last year or two, discussion of AI risk has become much more mainstream. On the one hand, I'd expect broad awareness of AI risk (especially within the AI and technology communities) to go a significant way towards reducing it, which is a terrific accomplishment. But on the other hand, as things become more mainstream they tend to get diluted, and dilution might wind up rendering AI risk reduction efforts ineffective.
What I mean by dilution is that, as an idea's "brand name" becomes more popular, influential people become aware of it, and they eventually start to relabel their own, pre-existing projects under the new "brand name". They have two big advantages: first, they probably have more money/status/fame than the people who invented the brand name; and second, their pre-existing projects are probably much more mainstream, and therefore easier to understand/believe in.
We saw this happen with the brand 'nanotechnology', which originally meant Drexlerian molecular manufacturing, and has since been repurposed to mean 'any technology with nano-scale features', with Drexler's original research program becoming an insignificant side branch. And we actually saw it happen twice with the brand 'Singularity' - originally it meant smarter-than-human intelligence; then it came to be about Moore's Law and exponential growth of technology along certain metrics; and now Singularity University has repurposed the word with the goal of making it as unobjectionable as possible ("provid[ing] educational programs, innovative partnerships and a startup accelerator to help individuals, businesses, institutions, investors, NGOs and governments understand cutting-edge technologies, and how to utilize these technologies to positively impact billions of people.").
So far, this hasn't that been that big of a problem with AI, because Superintelligence and the other key sources explain pretty clearly that existential and other catastrophic risk is an overriding concern. But, for example, a few weeks ago there was a Google-wide talk on DeepMind and the power of general AI technology. The audience, not being familiar with issues of x-risk, mostly asked questions about (for example) making sure that DeepMind wasn't used for military applications, which might be a good idea but won't do much to abate the catastrophic risk problem. I find it very easy to imagine Google creating an "AI risk board" so that all arguments about potential danger are referred to the board, and all evidence of potential danger is used as support by the board to gain more influence within Google, but where the board is controlled by academic bioethicists or others who don't think existential risk is a serious possibility.