Accelerating Future

Roko on the Friendly AI Problem at UKH+ Tuesday, Feb 2 2010 

friendly ai and videos Michael Anissimov 4:10 pm

Trackback URI

7 Responses to “Roko on the Friendly AI Problem at UKH+”

1. Panda says:

   February 2, 2010 at 5:56 pm

   I apologize, because I intend to focus on criticism of how Mr. Mijic presented his theories. I am not doing this to attempt an ad hominem against him or to evade Mr. Mijic’s substantive points. I express no opinion on those. I do this because I think futurists cannot afford to look sloppy, when they are already fighting an uphill battle. I have no doubt that futurism can be done in a very rigorous way. I have no doubt that Mr. Mijic can intelligently and eloquently do it.

   My criticisms are as follows:

   1. In this clip, Mr. Mijic has only one citation. When he actually appeals to experts (“experts differ”), he offers no citations. He discusses historical ways of thinking, which he can not have personally observed. Unsupported by a citation, he discusses how people once thought mice generated spontaneously. This example is not very helpful, but worse, without a citation, it makes Mr. Mijic look like he is making things up.

   2. I would also advise Mr. Mijic to not talk about things he is not trying to prove. Next time he should say, “let’s assume X” instead of half-defending X. His half-defenses often lack rigor and make X appear hard to believe. This hurts his credibility for his real points.

   Respectfully, Panda.

   Reply
2. Roko says:

   February 3, 2010 at 7:49 am

   My one regret with this talk was not carefully distinguishing between “CEV” and “CEV-like proposals”, as what I actually presented on a slide named “CEV” was not, in fact, CEV.

   Thanks for posting.

   Reply
3. nadc says:

   February 3, 2010 at 4:23 pm

   This was a great introductory talk to Friendly AI. I thoroughly enjoyed it!

   Reply
4. Sulfur says:

   February 3, 2010 at 6:14 pm

   I’ve watched till 6:54, and after 4th facepalm I decided to stop. That wasn’t serious presentation, I hope, but if he was trying to explain problems of friendly AI to laymans-he failed. Because even a layman will see that logic of (beginning of) presentation was based on wishfull thinking and handwaving. Starting from definition of inteligence -it can be easily applied to an electron; to the statement of 50/50 chance of AI emergence within century – he shouldn’t waste precious time on that which he couldn’t proove. What I can expect after that? oh yes, assumption of ever-expanding inteligence as a scientific fact, not philosophical construct. Because guys from “real” science are not doing anything connected to this “dirty” humanities, they are in realm of fact… He would be able to avoid all of that, if he would listen to Panda’s advices from the first comment here. That IF we assume that inteligence is X, and WHEN we will assume that AI will be created within 100 years from now, and IF artificial inteligence can be expanded beyond human capacity, THEN we have Y. But in such scenario he wouldn’t be able to pose as a serious scientist, a futurist with insight of things to come… I’m sorry for attitude, but this was quite dissapointment. I’m going to look trough it some another time, so thanks for posting anyway. I hope that the other parts are more solid, since beginning isn’t.

   Reply
5. MetaSeer says:

   February 5, 2010 at 4:04 am

   It’s not called pioneering research for no reason.

   Reply
6. Pilot Pirx says:

   February 5, 2010 at 8:13 am

   To reiterate: a requirement for friendly AI: friendly humans.

   Reply
7. Lorelei Wattigny says:

   January 20, 2012 at 9:24 pm

   Hi, i think that i saw you visited my web site so i came to ¡°return the favor¡±.I’m attempting to find things to improve my site!I suppose its ok to use a few of your ideas!! Just want to say your article is as astonishing. The clarity in your post is just excellent and i can assume you are an expert on this subject. Fine with your permission allow me to grab your RSS feed to keep updated with forthcoming post.

   Reply