Japanese Scientists Unveil Thinking, Learning Robot

[Jaime_Wolf]

It's not what I define as an artificial intelligence, per se, but I applaude the achievement. It is a step forward. My thing about the AI term is that AI should be more than just seek info and use info. It must formulate its own thoughts.

(To which, I refer back to another time where I jokingly stated that a proper AI should say something like "Fuck this, I'm off to Vegas." as proof of original thinking.)

My concern: It uses the internet and may not have the ability to determine what is fact and what is fiction. There are many volitile ways that things can go wrong with that, but here's a safey-yikes sort of one. If a robot were to decide to make tea and find a reference to Douglas Adams' Restaurant at the End of the Universe, in which Arthur Dent befuddles a computer with tea instructions that it cannot meet... Well, you see what I mean.

Humans can't magically determine fact from fiction either. It's an impossible task. That said, they suggested that it use the internet to learn from other robots with a similar AI scheme, so the Douglas Adams situation would never arise. And robust AI would likely be just as good at telling fact from fiction as humans. So just as a human is unlikely to mistake the Douglas Adams instructions for reasonable instructions, so too would the AI.

Also, you should think for a moment what "original thought" actually means. Seeking and using information isn't just AI, it's a remarkably concise definition of intelligence in general. That's exactly what humans do. Your thought thought about "Fuck this, I'm off to Vegas." is actually a great one, but not because it shows original thinking. What it would show is a robot with what amount to emotions reasoning rationally based on those emotions. People don't just come up with ideas like "Fuck this, I'm off to Vegas." out of the aether - it's a rational decision based on a belief that going to Vegas is more personally worthwhile than continuing the task at hand (based on information about Vegas, information about the task at hand, and information about personal satisfaction).

Neither machines nor people can reason from nothing. The thing that we call an "original thought" is really just a novel combination of information giving rise to a seemingly improbable piece of reasoning. If it isn't that what the fuck is it? Where does it come from? How did you come up with the idea if not from previous experiences and the ideas built off of them?

Ah well. Wouldn't be the first time I was wrong AND right at the same time. Still, it is a thing I wonder, regarding whether the robot will fair better or worse than humans in its learning curve.

I would be incredibly surprised if there were ever any kind of AI that wasn't a more or less exact replication of human intelligence (which is to say, human brain function). Frankly, it's hard to see how you could really do better in terms of engineering.

Probably the most interesting differences would be the contents of the thoughts rather than the mechanism underyling the intelligence. We're remarkably limited in how fast we can transmit information to one another, which limits the speed at which we can build new ideas out of the combination of those ideas and our own existing ideas.

In short, I don't see the intelligence being any different, but I definitely see the I/O devices being much more efficient and it's hard to say what that would mean. How would ideas evolve if they could be transmitted between brains effortlessly and instantaneously? It's very hard to know.

 

[Summerstorm]

Would be funny to see what happens if the robot learns (through the internet) that "orange plastic grain is not water"... "Humans are 73% water"... "Scientists are human"... "Your programme/engineers are scientists"...

"Robot could you fill this cup with water?"

"YES, my watery MASTER!"

 

[Evilsanta]

I just hope that no scientist on that said "How could anything possible go wrong?"

If someone did...Then we are all screwed.

 

[Cid Silverwing]

Why always Japan with the technological leaps?

They're a fuckton more disciplined than us lazy-ass Westerners. Comes natural to a nation that gave birth to Samurai.

I couldn't stop thinking of R.O.B for the NES (or was it SNES?) while watching this. It fascinated me to no end. Soon we'll have robot servants.

 

[Bobbity]

God damn, I still can't get used to the new Youtube player.

 

[Jakub324]

Why always Japan with the technological leaps?

They're a fuckton more disciplined than us lazy-ass Westerners. Comes natural to a nation that gave birth to Samurai.

I couldn't stop thinking of R.O.B for the NES (or was it SNES?) while watching this. It fascinated me to no end. Soon we'll have robot servants.

Can I remind you that we gave birth to the Puritan? Looking forward to being able to say: "H4! Get me a drink."
"Yes, master." I might even get a few years service before he strangles me in my sleep during the robot uprising.

 

[Danish_4116]

"SOINN obtains information from other sources including the internet as well as other robot's experiences and knowledge. In this way, it gradually becomes smarter."

Yep. That's not a recipe for disaster.

I can imagine it now.

"What is fairness?"

I was thinking more the end of Terminator 3.

John Connor tries to destroy Skynet's mainframe only to discover that it had uploaded itself to every computer in the world via the internet making it impossible to isolate and destroy when it eventually turned on humanity.

So when the day comes that SOINN asks "What is fairness?" the scientists will freak out and try to pull the plug, but will only post-pone the robot uprising as SOINN will already be living in computers all over the world :S

I'm going to take my laptop for a swim now.

 

[TheAmazingHobo]

Plausible patterns? An AI that can make simple decisions is a far cry from sentience which is itself a far cry from robot apocalypse. Our level of AI is still so pitiful that predicting the end is a gross jump forward. Robots are still incredibly stupid, as anybody who has studied AI can tell you. They are definitely getting better but they'll still a far cry from anything warranting the kind of response that they get on these threads. The overreaction is a bad joke. It was funny a few times but after hearing this joke over a dozen times on this website I've lost my humor. I feel like I'm in a room full of doomsayers talking about something they know nothing about. For as little as I know about AI, I still know that the kind of outlandish things people are saying is just as cray as saying "she turned me into a newt." It's not predict the future based on trends, its drawing the line from current trends to the future to fit your own views. It's far closer to calling out a witch then predicting the future. Because of all your silly jokes about Skynet I can't help but see these posts as being anything more then silly. Super silly (us).

I would like to echo the sentiment put forward by my distinguished colleague and maybe also hug him a little.
Every thread about anything robot related always degenerates into unbelievably repetetive jokes or discussions about sentient and souls, as if that had anything to do with the news (for those playing at home, it doesn´t).

Personally, I blame the op.
From how he phrased it, it is difficult NOT to think that we are about half an hour away from unleashing the Geth.
Though I guess I can understand. Anyone wants to know what is the actually interesting thing about SOINN ? It offers a decent approach to the Stability-Plasticity Dillema and its robustness with regard to degenerated and noisy input.
That´s not the stuff that headlines are made from.

 

[UnknownGunslinger]

Hooray, we may have finally developed a robot with the intellectual capacities of an insect!
Why does it speak like a girl is something I prefer to not know the answer to Japan

 

[Meight08]

I think your silly fun detectors are turned off.

OT:this is interesting but i hope they will pu some guidelines in place

 

[Fanta Grape]

Oh, do these are the motherfucking spam bots that have been managing to get past the captcha codes. Touché, Japan...

 

[WabbitTwacks]

So what happens if you ask the robot to kill somebody and then it uses the hive mind to teach other robots to do it?

 

[Metalrocks]

some technologies are alright and good, but this i think is not alright at all. so much for i-robot and terminator. hope i will be dead when the war starts.
who knows. mass effect could turn out to be real.

 

[Use_Imagination_here]

Think that's impressive?

http://www.youtube.com/watch?v=RNKLuXUh3M4&feature=related

That is a robot telling a man he "loves" he is afraid of leaving for another country. The future's knocking on the door.

To be honest I think a robot actually learning something is more impressive. Sure, the mere mimicry of human expression like the Hanson robots do will have them pass the Turing Test, but it doesn't mean the robots are actually feeling for themselves.
I know you're not implying any different, but you're wording makes me feel the need to reiterate that this robot has been told to say that he is afraid by extracting information from humanly-compiled database, not because he's drawing a conclusion from a neurological response.

No, that robot hasn't been told to do ANYTHING since it was created. Everything it said in that video it said because of interaction with his creator and learning.

 

[anthony87]

Wow, that article was bad. Where did you even find that? It's style was all over the place. Why would the writer even suggest about avoiding that "cliché", when he's then going to go and say something like "you'll need that innate hope when the metal ones come for you"?
This just reads like blatent alarmism to me and look what it's fuelling:

I'm sorry, Dave. I'm afraid I can't do that.

Oh lord.. what are we doing..

and

Fucking hell Japan, you'll be the death of us all.

Look, I know it's the most reliable survival tactic to fear everything that's unfamiliar, but it's part of beign a human that we can use knowledge to counteract those primal fears. Yes, robots could potentially be turned into killing machines, but if humans managed to become social animals of which a staggeringly high percentage are docile (only 2.84 percent of death is caused by violence, which is by far the lowest of any carnivorous species) and we are the ones who have to teach robots to think, then we can also teach them to be agreeable creatures.
If they do turn into death machines, then that's won't be because robots themselves will be inherently evil or selfish, but because humans did that to them. I think by far the biggest undermined threat here is that PEOPLE SUCK. Not inherently of course, but it's the who do that have malicious intent and it would be those who would teach that to robots.
And furthermore, other revelations in science also caused social unrest, such as the discovery of nuclear fission and genetic engineering, but those never caused an apocolypse and probably never will (no matter how much fiction writers would have you believe that). This is also why I disagree with Jehovah's witnesses so much, because for all the doing good for humanity and listening to what J.C. said about loving each other, their insistence that there will be global ruination during the time that we are alive, simply isn't being realistic, considering that we've been around for millenia and life itself has been around for millions of years.
So, with that in mind, just sleep easy and carry on with life, because it's all going to be OK. But, I'm thirsty after all of this, so now I'm going to go and pour myself a nice cool cup of water.

Source: [link]http://en.wikipedia.org/wiki/List_of_causes_of_death_by_rate[/link]

My apologies for having a bit of fun with a post....?

I even smiled as I was typing it, of all the nerve eh?

 

[Anti Nudist Cupcake]

Ok so wait, a robot that can exchange knowledge with other robots over the internet which is also a massive network?

Where have we seen that before... And what where the consequences of that ?

Spoiler

Oh yeah, that's what. They learned to shoot at things that contain organs.

 

[vid87]

So there's your fail-safe: before they all go rogue, program them to think that all water is actually a bunch of pebbles. Then, while they're chasing us, lead them into a river.

*Dusts off hands* Problem solved

 

[Atmos Duality]

Two posts in...and there's the boring Skynet reference.
*sigh*

It's the sort of thing that just sinks my interest in a topic I should be interested in.

Why always Japan with the technological leaps?

It's not always just Japan; but we tend to over-emphasize Japan's scientific contributions in publication.

 

[k-ossuburb]

Why always Japan with the technological leaps?

I dunno, DARPA in Boston made a pretty kick-ass robot. It's also creepy in how "alive" it looks, it almost behaves like a living quadruped.