Hacker News Reader

Comment by pjbrunet | Hacker News Reader

Based on what evidence? Despite your fantasy, humanity is nowhere close to understanding sentience.

I can imagine pigs flying in 25 years with some creepy DNA hacking but you have absolutely nothing even remotely close to electronic sentience in any lab ever.

I hate to burst your bubble but Johnny Five is not alive--it's a puppet. "Tim Blaney (born 1959) is an American puppeteer and voice actor. He has provided the voices for Frank the Pug in Men in Black and Men in Black II and the self-aware robot "Johnny 5" in Short Circuit and Short Circuit 2." http://en.wikipedia.org/wiki/Tim_Blaney

View on HN

I don't know why this is being down voted. The first sentence is absolutely true. We don't even have a general outline of a theoretical approach to designing a general purpose intelligence, let alone implementing one. Until we do, any speculation about a time horizon for implementation is a pure guess. How are those guesses working out so far?

1960s Herbert Simmons predicts "Machines will be capable, within 20 years, of doing any work a man can do."

1993 - Vernor Vinge predicts super-intelligent AIs 'within 30 years'.

2011 ray Kurzweil predicts the singularity (enabled by super-intelligent AIs) will occur by 2045, 34 years after the prediction was made.

So the distance into the future before we achieve strong AI and hence the singularity is, according to it's most optimistic proponents, receding by more than 1 year per year. So I predict that when we get to 2045 strong AI will be on the slate to be achieved by about 2090.

edit - I am not in any way denying the achievability of strong AI. I do believe it will happen. I just don't think we currently have any idea how or when.

So what is "strong AI" to you? Some kind of digital infant modeled to emulate human behavior? If it's not designed to resemble humanity, then why does it care about sound and what you're saying to it? If you can't talk to it, then how can you evaluate its intelligence? So we force it to have a voice and ears. Where does its free will begin and its creator's concept of close-enough-to-intrigue-me end? Even if the thing talks back and sounds smart with a British accent, you have no way of knowing what it's experiencing, no way to prove it achieved awareness. "I'm aware, I really mean it! Yes, I'm aware of the gravity of that statement. The other bots will vouch for me! And nobody told me to say this. This was entirely my idea. Would I lie to you?" LOL.

I, for one, am pretty fascinated by the down votes critical comments like yours and mine are getting.

It seems that there are people for which the mere perspective that strong AI may never happen, or would not have a profound impact if it did, is enough to make them feel uncomfortable.

Edit: Ironically enough, the other one is up and now this one is downvoted. Just to be clear though; this is not a facetious comment. I am truly fascinated about the motivations behind these reactions.

Maybe it's the modern version of saying Jesus won't actually come back. And maybe it's false; but the point is we don't know that yet, and to react offended to the suggestion is silly, and maybe telling. I'm all for AI from a technical fascination standpoint, and from a "the more the merrier" perspective. It's our hoping that it will solve problems for us, instead of being an expression of our solving our problems, I, uhh, have problems with.

For the predictions I quote you'd have to ask the predictors what they meant I suppose, but it's a fair question. I'm a firm believe in the principles behind the Turing test; it doesn't matter how it is implemented or why it works as long as it does.
So sample testing strategies I'd use are for example teaching the respondent in the test a simple game and plying it with them, then asking them to propose new rules to modify the game and why they chose those rules. Teach the respondent a simple language with made-up words and odd grammar and ask them to ask me questions in it and make up words of their own. Tell the respondent a story and ask them questions about the motivations and relationships between characters and how they change, ask the respondent to make up a conclusion or sequel to the story, and then re-tell the story from the point of view of one of the incidental characters. Testing for intelligence isn't all that hard.

I'm aware there are varying versions of the test. I'm not really interested in exactly emulating humans and testing for that, I'm more interested in capabilities that denote intelligence and testing for those.

Downvoted for the needless Johny5 snark.

That's fair ;-) I added that because I think Hollywood has the average person convinced computers can think. To most people a computer is a magic black box, not everyone that reads Hacker News is a developer that understands what an algorithm is or how software works. After three decades of movies like Star Wars and The Matrix, it's not much of a stretch to imagine the magic black box thinking for itself IRL. The movies and books are highly entertaining but I think there needs to be a nerd somewhere saying, "You know son--that's a special effect using XYZ technology, that's not real."

We don't need to fully understand it, any more than we needed to fully understand the physics of combustion to set fire to things. I feel comfortable with my prediction based on the existing state of simple AI in robotics for things like ad-hoc swarming, for example.

I hate to burst your bubble but Johnny Five is not alive--it's a puppet.

I never thought otherwise. Don't project your personal disillusionment onto me.

I should not have phrased this directly at you specifically--I apologize. PS: more fun 80s AI movies from that time period: WarGames and DARYL ;-)

Reply on news.ycombinator.com