3

lukeprog comments on Looking at how Superforecasting might improve some EA projects response to Superintelligence - Effective Altruism Forum

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (4)

You are viewing a single comment's thread.

Comment author: lukeprog 30 August 2017 10:37:43PM 2 points [-]

Without taking the time to reply to the post as a whole, a few things to be aware of…

Efforts to Improve the Accuracy of Our Judgments and Forecasts

Tetlock forecasting grants 1 and 2

What Do We Know about AI Timelines?

Some AI forecasting grants: 1, 2, 3.

Comment author: WillPearson 31 August 2017 08:12:40AM 0 points [-]

Thanks for the links. It would have been nice to have got them when I emailed OPP a few days ago with a draft of this article.

I look forward to seeing the fruits of "Making Conversations Smarter, Faster"

I'm going to dig into the AI timeline stuff, but from what I have seen from similar things, there is an inferential step missing. The question is "Will HLMI (of any technology) might happen with probability X by Y" and the action is then "we should invest in most of the money in a community for machine learning people and people working on AI safety for machine learning". I think worth asking the question, "Do you expect HLMI to come from X technology". If you want to invest lots in that class of technology.

Rodney Brooks has an interesting blog about the future of robotics and AI. Worth keeping an eye on as a dissenter, and might be an example of someone who has said we will have intelligent agents by 2050, but doesn't think it will be current ML.