Japanese scientists develop a 'thinking' robot

[Humodour]

http://www.google.com/hostednews/afp/artic...8b87cd6a8f7.2d1

 

****ING AWESOME.

 

I think some of the Obsidian devs should take a look at this. It's a shame we can't tag people in forum posts like we can on Facebook.

[Serrano]

If it learns largely from the internet that that would be a bad thing surely. It would basically be a new generation of l33t gamers only they'd be super intelligent about how they insult and greif you.

[Orogun01]

Well gentlemen Skynet is up, it has been a pleasure knowing you.

[Gorth]

Skynets last words to John Connor before it lmind-wipes him and turns him against his fellow humans: "r00fles!"

[Lexx]

Why I must login to google to read the content of the link?

[Rosbjerg]

Why I must login to google to read the content of the link?

 

I can read it without login in.. perhaps they really want your data?

[Morgoth]

People are way too optimistic when it comes to AI development. Current robots have barely the intelligence of a mouse, and it will probably take another 50 years to have robots with the intelligence of a cat or dog.

[Gorgon]

Well of course. we are building artificial intelligence from the ground up. We wouldn't be doing that if it was autonomously learning.

[Humodour]

Some more gems for those interested in the field:

 

http://www.physorg.com/news/2011-10-boston...tary-robot.html

http://www.physorg.com/news/2011-10-mabel-...gged-robot.html

http://www.asiaone.com/News/Latest+News/Sc...720-290039.html

 

People are way too optimistic when it comes to AI development. Current robots have barely the intelligence of a mouse, and it will probably take another 50 years to have robots with the intelligence of a cat or dog.

 

Um. You don't understand the field. It depends entirely on what your definition of 'robot' or 'AI' is. We have robots and AI's right now which are extremely good (human-level or better) at specific tasks or groups of tasks. We've got computers which combine thousands of different AI algorithms into one and use probability theory to make decisions, much like a human brain (IBM's Watson), we've got things like this above, and we've got artificial neural network robots, we've got real neural network robots (neuron cultures), we've got robots that understand human speech, we've got robots that can perform image recognition as well as humans... robotics is now an extremely diverse multi-billion dollar field which continues to grow at a spectacular pace. Probably the biggest limiting factor for robots and AI is NOT modern technology or algorithms, but computing power. And computing power is still increasing faster than exponentially. Does any of the above qualify computers for having sentience yet? Certainly not. Does it need to? Certainly not. Will it in the near future? Most likely.

 

Always the pessimist, Morgoth. And often fairly off the mark, to boot.

[Morgoth]

Um. You don't understand the field.

 

We have robots and AI's right now which are extremely good (human-level or better)

 

Wow, you truly are a clueless idiot.

[Gorgon]

Stop with the name calling.

 

Anyway just because computers have been able to beat the best human chess players for a while now, that doesn't make them intelligent. AIs all lack the social aspect which requires knowledge of human nature and empathy. That requires being human, or perhaps the memory banks of a human being, but since we can't transfer consciousness just yet it's all academic.

[Morgoth]

Always the pessimist, Morgoth.

And your perpetual optimism is actually caused due to a brain disfunction.

[Rosbjerg]

Hmm, perhaps we should focus on discussing the topic rather than the mental abilities of our fellow posters?

[Malcador]

Wonder what it is the Japanese have with robots. The Alpha Dog is neat from a controls systems POV, wonder how well that'd do in military service.

[Gorgon]

Humans are really cheap by comparison, I guess that's why Japan doesn't have a Mecha army yet.

[Morgoth]

Japans has a rapidly aging population, and as opposed to us fancy Europeans who let cheap labor care for the eldery, Japan with practically no immigration has to be more...innovative, hence the robot army to care for the eldery. I like the novelity, but it ain't gonna work.

[Humodour]

Always the pessimist, Morgoth.

And your perpetual optimism is actually caused due to a brain disfunction.

 

Look, first off mate, you can quit with the mean-spirited and rude insults and name-calling ok? Try some civility for a change. Do you treat everybody in your life like you treat people on here?

 

And maybe instead of posting a link to an incorrect and poorly worded journalist's perspective on a scientific study, you could go and read the study yourself?

 

http://www.nature.com/neuro/journal/vaop/n...ll/nn.2949.html

 

Morgoth, you called me an 'clueless idiot' because I pointed out that there are areas of intelligence where computers outperform humans now. This was arrogant and ignorant of you, and it makes you look like a buffoon, because what I said was correct. These are commonly called expert systems and are already widely used commercially. They outperform humans on specific tasks. The challenges is integrating them together so that you've got something which can outperform humans on multiple tasks.

 

Perhaps the most noteworthy AI is IBM's Watson because it is capable of using true learning methods like neural networks and genetic algorithms to process and integrate the information provided to it by 1000's of competing expert systems and AI algorithms (some of which are themselves true learning algorithms). In doing so it consistently beats every human player at a game called Jeopardy (you might want to look into the rules of that game if that doesn't impress you), and is now being used to diagnose medical disorders because it can process (and access) more information than doctors: http://spectrum.ieee.org/riskfactor/biomed...-ibms-dr-watson

 

We also have organic neural nets (which are real neurons from a rat brain which are essentially encouraged to form links with eachother in a petri dish - calling it an artificial brain would be accurate) which are capable of some surprising feats: http://singularityhub.com/2010/10/06/video...moving-forward/

 

Then there is the Blue Brain project being funded by the Swiss government to run an entire mammal brain by taking cortical slices and translating them into a computer (i.e. including gene regulation, neural network, neurotransmitters, the whole gamut - rather than just trying to simulate the neural network). The only limiting factor to this project is supercomputing power... which is increasing faster than exponentially.

 

http://bluebrain.epfl.ch/

 

Morgoth, I don't think it is optimism to expect the current pace of progress in robotics and AI to continue into the foreseeable future. If anything, I think that is a slightly pessimistic assessment (computing power has started violating Moore's law by advancing faster than it, as one example of the factors propelling this field forward... another factor is the similarly lightning-fast pace of materials science research, and another is the entrance of scientists and governments in India and China into the field - more hands and funding make light work). But this slightly pessimistic assessment based on the pace of CURRENT progress in the field is an assessment I run with because it provides a nice floor to work with when projecting future developments.

[Humodour]

Japans has a rapidly aging population, and as opposed to us fancy Europeans who let cheap labor care for the eldery, Japan with practically no immigration has to be more...innovative, hence the robot army to care for the eldery.

 

3 of the other top 4 biggest robotics markets, commercially, are South Korea, Germany, and the USA. So, what are your excuses for these countries?

 

I like the novelity, but it ain't gonna work.

 

Why not? Maybe you should let the Japanese government know. I'm sure they'd appreciate your expertise in robotics.

[Walsingham]

Purely by way of devil's advocate, what if increases in computing power, and manifold applications for 'dumb' computers will actually fracture and dissolve efforst to achieve true AI?

 

After all, when it comes right down to it, what's the pointof having true AI? What would it actually do? Clean the house? Plan my accounts? Fight wars?I can have a human do these things for next to nothing compared with the cost of AI.

[Humodour]

Purely by way of devil's advocate, what if increases in computing power, and manifold applications for 'dumb' computers will actually fracture and dissolve efforst to achieve true AI?

 

What I expect is that AI will, over a long period of time, slowly replace the algorithms for things which we've always done or used, and people will barely give it a second glance, because it happened so slowly (relative to humans - i.e. over a period of years). Look at Google Translate or Google search. Some pretty sophisticated machine learning algorithms (AI) at work there, but nobody really notices that. Certainly a far cry from the dumb Google of 1999 that I remember (although yes, it was still the best search engine at the time).

 

After all, when it comes right down to it, what's the pointof having true AI? What would it actually do? Clean the house? Plan my accounts? Fight wars?

 

A great response I can think of off the top of my head is science. Something that could do science for us day in, day out, without losing its knowledge upon death, without needing to train replacements for, at a speed that can be increased indefinitely with increased computing power (which comes from... more science) would be pretty ****ing useful.

 

It's the type of thing which applications are found for once it is built... like LASER or the Internet.

 

I can have a human do these things for next to nothing compared with the cost of AI.

 

But what is the cost of AI? The software cost is nothing (besides the initial investment). The hardware costs... well, I'd suggest that in 20 years time the supercomputers of today will be pretty cheap. That has been the trend for the last half a century or so. Look at your smartphone.

 

There are things we won't want AI/robots to do for us, because those are human things which we will cherish as part of what it means to be human. But there are things we will.

[Morgoth]

3 of the other top 4 biggest robotics markets, commercially, are South Korea, Germany, and the USA. So, what are your excuses for these countries?

That those countries already have/or soon will have a rapidly growing old population?

 

Why not? Maybe you should let the Japanese government know. I'm sure they'd appreciate your expertise in robotics.

As I said before, a) not enough processing power yet to have a robot do flexible and autonomous work on "his" own yet, unless you get a human brain to work with the signaling (which would cause a lot of ethic questions and bruarahah) and b) right now it seems challenges pertaining to mechatronics/electrical engineering/battery life etc. make it almost impossible to let a robot do all the physical work in a satisfying way. No point to carry grand-grandma up the stairs, put her on the toilet, watch that nothing goes wrong, put her up, wipe her ass, carry her down, cook lunch, etc. when the stupid thing already fails at getting up the stairs without loosing balance or another gazillion issues. Not in this decade...but perhaps still in this century? (OMG I'm now an optimist too).

[Walsingham]

Your concept of a Science! robot sounds good. Except that, as I underlined in my original question, science doesn't occur ina vacuum. Any scientifically observed phenomenon only becomes useful in human terms when it married to a human individual or organisational need.

 

Leaving aside the issue of who'd want to finance an AI that was enquiring at random, just consider the raw size of the volume of ignorance it'd be attacking. An almost infinite variety of permutations in experimental conditions and subjects.

 

Further I think you're underestimating the actual value of having humans doing experimental work. Yes, they get bored and distracted by pretty first year students. But they also spot 'extraneous' data that is useful. Like the way a garbage lorry is outside whenever they get weird data. Or that the green mold which keeps upsetting their experiments is actually penicillin.

 

The reason I'm belaboring this point is that you're an educated guy with a lot of energy. The sooner you understand that everything has to have a point, and people invested in it, the sooner you'll be able to turn some of that energy into useful work.

[Humodour]

Your concept of a Science! robot sounds good. Except that, as I underlined in my original question, science doesn't occur ina vacuum. Any scientifically observed phenomenon only becomes useful in human terms when it married to a human individual or organisational need.

 

Leaving aside the issue of who'd want to finance an AI that was enquiring at random, just consider the raw size of the volume of ignorance it'd be attacking. An almost infinite variety of permutations in experimental conditions and subjects.

 

Further I think you're underestimating the actual value of having humans doing experimental work. Yes, they get bored and distracted by pretty first year students. But they also spot 'extraneous' data that is useful. Like the way a garbage lorry is outside whenever they get weird data. Or that the green mold which keeps upsetting their experiments is actually penicillin.

 

The reason I'm belaboring this point is that you're an educated guy with a lot of energy. The sooner you understand that everything has to have a point, and people invested in it, the sooner you'll be able to turn some of that energy into useful work.

 

Oh really?

 

http://www.physorg.com/news/2011-10-robot-...ex-problem.html

http://www.itnews.com.au/News/276700,ibm-e...-computing.aspx

http://blogs.hbr.org/cs/2011/10/apples_sir...volutionar.html

 

If you don't understand what you're talking about, keep quiet, Wals. Same goes for you, Morgoth. You two are to artificial intelligence what Luddites were to the industrial revolution.

[Walsingham]

 

Indeed. And until you stop posting me links to stories about expert systems rather than the kind of AI you've been waffling about, maybe it is YOU who should keep quiet.

[Morgoth]

If you don't understand what you're talking about

Yeah, right. Reading science articles all day makes you understand things so much better than us plebs.