Eliezer Yudkowsky on "Open Problems in Friendly Artificial Intelligence" at Singularity Summit 2011

The Singularity Summit 2011 was a TED-style two-day event at the historic 92nd Street Y in New York City. The next event will take place in San Francisco, on October 13 & 14, 2012. For more information, visit:
http://www.singularitysummit.com

• Category

  Science & Technology

[AntsanParcher]

I was told not to look into the comments. I didn't listen. So I suffered. Well, at least now I know *why* I shouldn't have looked into the comments, which is kinda nice.

[AwsmAdm]

I was told to not look at the comments, and so far I've only read this one. So far so good; wish me luck.

[Jotto999]

I wasn't told anything, but I've been browsing YouTube for plenty long enough to know in advance what to expect from the comments, and I will quietly take your and AntsanParcher's lead.

[polyopulis]

Because it would have to value its advancement over helping humankind in the first place.

[XetXetable]

It would be easier to move itself to isolation from humans (which isn't difficult to do, even if it stays on earth) than to destroy humans. You present a false dichotomy.

Additionally, if we don't know and don't have any way to assign a probability to a given outcome, there's not much reason to speculate on the details of any, unless you're looking to write a sci-fi book or something.

[AntsanParcher]

Sorry I am still commenting the comments instead of watching the video itself and commenting on that but it seems, I am unable to learn.

I saw that there was an answer to my comment and looked, what it was (and what I originally wrote). Then I got curious what those comments were... Maybe I will listen to myself the next time.

[lordjavathe3rd]

I don't like saying when things will happen, I'm only concerned about if they can happen and what are the prerequisites. I want to know if the problem is well defined, if not then we can define it. I want to know if we have the hardware required or how do we know what is required, if we don't have it, we'll make it. I want to know if current methods are old and tired, if they are then I will make new ones. Conscious AI is no concern of mine, nor will it ever be.

[dizekat]

Rationality. Why study or work when you can talk people into paying you for doing work that you are not even remotely qualified to do? That would be irrational. (Dropping out of school wouldn't have been so bad if he created some obviously useful or cool things as he 'educated himself')

[otonanoC]

Yudkowsky is a bowl of fruity fruit.

[Aaron C]

Whos to say the AI would choose to help mankind instead of seeing that we are in the way of their advacement; thus squashing us as mere roaches?