Archived

This topic is now archived and is closed to further replies.

Pixel_Sticks

Terminator 3 inspired question, err sort of...

Recommended Posts

Given the current state of AI used in Robots how long do you think it will be before a complete autonomous robot is developed that can actually do something useful? I watched a program on discovery channel a few weeks ago on the subject and one guy said the robots today have the IQ of a retarded cockroach. Another computer scientist said that he doesnt believe that robots will ever be able to get to the point of autonomous machines who can learn and maintane themselves as seen in the movies. What are your feeling on the subject? Termintor 3 inspired me to ask this question but I have always been interested. ~PS

Share this post


Link to post
Share on other sites
The problem is that a ton of people who would want to experiment with AI don't have money to build a real-world robot. Without much outside input, a robot is pretty worthless.

I saw T3 today. Pretty cool.

[edited by - Nypyren on July 3, 2003 3:46:47 AM]

Share this post


Link to post
Share on other sites
Scientists have to work harder on the boddies of robots...
Any robot today has a full mechanical body, we need robots that have rubber muscles and actual synthetic skin, so that other AI scientists might build something on top of that...
The brain will be easy (flaming material here). I know what i say because i do AI research myself. Humans will:
- stumble on the answer (has happen before)
- brute force it (wait 100 years or so, and any PDA will run AI code, do a "seti" on an artificial brain and voila!)
- actually do the work (expect amazing stuff in 12-18 years)

AI research is very good a coming up with very good specific ways of doing one task, like voice recognition, visual recognition, path finding, etc.
Real AI has trouble coming up with a solution that will be generalistic enough to encompass it all...

My $0.02 anyway... ;D

[Hugo Ferreira][Positronic Dreams]
I Am Digerati.

Share this post


Link to post
Share on other sites
There are many autonomous robots that do useful things, they just don''t look like people, walk around on two legs, or start conversations.

Assembly plant robots are autonomous (once you get them going). They use many complex algorithims to make sure that they are exactly where they should be, and are constantly adjusting themselfs.

Roomba, the robot vacuum cleaner, is usefu and fully autonomous. It still wants you to empty it''s dust bin and plug it in for recharging, and needs you to turn it on, but thats about it.

Stinger missles are autonomous to a certain degree, as are Exocet curise missles. Fire and forget.

The computers in your car run all by their lonesome without any human help and do a fine job of keeping everything running smoothly.

Telephone swithes do some pretty nifty autonomous routing and load-balancing.

Penut processing plants can be fully autonomous factories-- humans only show up for maintenence, as the operating conditions are ''hazerdous'' to humans.

There are tons of useful autonomous robots out there.

Cheers,
Will

Share this post


Link to post
Share on other sites
I''ve made some preliminary designs for a useful robot but haven''t had money or time yet to work on it.

My goals are pretty simple to itemize but involve a lot of work to implement:
- Map and navigate a house.
- Open/shut doors and cabinets.
- Map shelf and cabinet geometry and recognize items.
- Pick up, cary, and put down light objects on any platform from 0 to 7 ft high.
- Respond to voice keyword commands using a hypertext-like interface. You access pages of text-to-speech by keywords and the robot uses a different pitch on the keywords so you know where the speech links are. I know...sounds silly but it''s functional.

The robot itself would look like a short R2D2 with a tall pole sticking up that manipulates an arm/gripper assembly like a crane.

Some applications:
- Fetching/Putting small stuff away
- Dusting/sweeping/mopping/polishing

Does anyone else have a million half-baked things like this they never seem to be able to make progress on?

Share this post


Link to post
Share on other sites
Guest Anonymous Poster
quote:
Original post by 5010
Does anyone else have a million half-baked things like this they never seem to be able to make progress on?


I do - though the secret is to throw away your 999,999 other ideas and concentrate on just one.


Back to topic - there''s not much hard evidence either way on this. It''s just a matter of what you want to believe really.

I don''t think anyone (university research, private companies) has aimed as high as you are suggesting and been willing to commit to it.


Share this post


Link to post
Share on other sites
quote:
Original post by 5010
Does anyone else have a million half-baked things like this they never seem to be able to make progress on?

i am waiting to win the lottery to begin the actual implementation...

Share this post


Link to post
Share on other sites
Well, in order to actually take over the world, robots will have to be able to:
-Repair themselves
-Know how to build new robots
-Know how to design and program new types of robots
-Use ingenuity (adapt to all situations)

That''s a hard job.

Share this post


Link to post
Share on other sites
Well, Zolcos, your post is where AI hits the fan, for all
the bad reasons.

Im going to show you a living being, that has for all preactical pusposes conquered the planet, and he cant do anyhting on your list, ants:
- an ant cannot repair itself: less energy is spent on "building" a new ant
- 99.9999999999% of all ants on this planet have never heard of sex, or reproduction...
- an ant is quite incapable of adapting to situations she wasn''t pre-programmed to. Yes, at first glance everybody is going to argue that ants are super-adaptable, and i agree, ants are, a single ant isnt...

AI is so many diferent things to so many diferent people, that is is almost mind-boggling. BTW, if you haven''t noticed, ants are my favorite animal, and their military model of operations is probably the best nature has to offer. Future martial robots should follow their modus operandis...

[Hugo Ferreira][Positronic Dreams]
I Am Digerati.

Share this post


Link to post
Share on other sites
quote:
Future martial robots should follow their modus operandis...




...then goodbye humans! insect politics is an oxymoron but something I love to throw around at work, cauz they aint got no politics.. watch the nature channel sometime, they are brutal >)

I fseek, therefore I fam.

Share this post


Link to post
Share on other sites
Guest Anonymous Poster
quote:
Original post by pentium3id
Im going to show you a living being, that has for all preactical pusposes conquered the planet, and he cant do anyhting on your list, ants:
- an ant cannot repair itself: less energy is spent on "building" a new ant
- 99.9999999999% of all ants on this planet have never heard of sex, or reproduction...
- an ant is quite incapable of adapting to situations she wasn''t pre-programmed to. Yes, at first glance everybody is going to argue that ants are super-adaptable, and i agree, ants are, a single ant isnt...



I don''t think your example of ants interferes at all with Zolcos ideas. You say an ant is intelligent but ants are. But Zolcos used the plural too I think the ant model fits well to the robot world - there won''t be a single robot type (like the generalised human) doing every task. It''s pointless and wasteful.

More likely there would be worker robots, robots what repaired/replaced worker robots, and then robots that controlled underlings (ie the "queen". Look at the Borg - it''s the way to go...

Share this post


Link to post
Share on other sites
I reckon the scariest "robots taking over the world" scenario is where microscopic nano-bots (bacteria bots?) reproduce endlessly and cover the world in gray goo. No intelligence needed, no repairing needed, just cheap easy reproduction and evolution. Mind you, imagine all the cool things we could do with "good" nanobots. They could fly around my loungeroom and change the channel for me.

I think theres a better chance in the next 20 years or so getting miniature swarm robots to work well than indivual superintelligent humanoid ones.

Share this post


Link to post
Share on other sites
Hmm...if you realy want to find out what is going to happpen to robotic socity in the next fiew years, I sugest reading a amazingly good book(yes, you will actualy have to do that think you learnd in schoool called ''read'') called Out of Control by Keven Kelly.

Even tho it was written in 1994, this book still contains lots of introsting informations about systems and robotics. And, many of his predictions 80% of them are happening right now as we speak.

The machine automotive system will be like a bee hive. However, dont mistake that the machines and programs will have a monarchy system. They will have a nural/democratic system. What that means is that each machine operates independently of its function, and only changes its functions when outside stimuly is imputed. Add a group of machines/programs, the machine world whould adapt like a super-organisum.

You see, even tho we call it the Queen Bee, this is just a name. A place holder, like our presidents. Her only job is to produce more babies, and make shure there genetics match there function for better success. The rest of the workers are actualy independet of each other, only adapting when certant information is suplyed. For example, if they run across a flower that is in polination, they will go over and sip its nector. The only time when they do socal status is when the vote on something. Yes, bees anctualy vote befor they go out of the hive. Thus, they retain there own induvidualty, while keeping the "super-organism" in balanced. Just like humans....

Anyway, just read Out of Control. Keven Kelly can explain it much better than I can! -.-;

Share this post


Link to post
Share on other sites
Guest Anonymous Poster
Woah there sidepocket. Find yerself a spell checker...

On that comment about grey goo - That is an idea thought up by the same kinda ppl that think GM crops will grow legs and start running into ppls houses to kick their pets. It is a gross exageration of what could possibly happen.

Share this post


Link to post
Share on other sites
And why should the robots take over the world? Why do you think that robots will think just like humans? S. Lem in his book "GOLEM XIV" (1981), writes about super computers, which becomed more inteligent than humans - but they don''t want to take over the world - they don''t care about humans, politics.

Share this post


Link to post
Share on other sites
quote:
Original post by Scrambles
I think theres a better chance in the next 20 years or so getting miniature swarm robots to work well than indivual superintelligent humanoid ones.


Then your government will insert this nanothings into your body when you are born and will monitor all you do. And if you do something against the law, the nanobots will automatically practice a lobotomy into you.

They will be able to make you stop when police shuts you to stop or even to die.

That will be the ultimate man-made freedom.

Share this post


Link to post
Share on other sites
Machines will NEVER EVER NEVER think like humans. Why? Well, for one, the human brain is not accurate enough to warrant spending billions to reproduce one. We forget, lie, cheat, steal, mis-interpret, exaggerate, make unfounded leaps of logic. Basically, we do everthing wrong when out into a context of a ''robot'' or ''machine''. Another reason, for a machine to think like a human, it would require all the sensory inputs we have. A robot would not be able to make inferences about items in its world they way we do unless it could ''see'' it the way we do. More than likely, a robot will have some senses above and beyond our scope, and others not at all. Robots would prolly have IR or UV seeing ability, or be able to hear sound in frequencies impossible for us. So it will see its environment completely different than we do. Also, for it to be human-like, it would require a human life span to obtain and filter and process the information. Human brain is theorized to be able to make 1 quadrillion calculations per second. Perhaps our brains potential actually hinders us, as we try to maximize our knowledge capacity, but until we can make a robot brain with that computing power, it will take ungodly amounts of time to build up neural development the way a human can. Perhaps specialization is the key and can overcome the current limitations of time and technology. As this was all started from a Terminator 3 inspired question, look at the terminators. Yes autonomous and self sufficient, but extremely specialized.

Share this post


Link to post
Share on other sites
pentium3id, ants do rule. I find it amazing that with all human technology we have now, we can put people on the moon, send probes to mars, know what a star 100billion light years away is made of, its size, its gravity, yet we cannot even come anywhere close to making a machine as efficient or small as the ant.

And I agree, if ants wanted to, they could annihilate mankind in a matter of days or weeks, but luckily for us, we are insignificant to them in their plan for universal domination... =)

Share this post


Link to post
Share on other sites
Stonicus I made the exact same observation while I was at the park with my girlfriend. She asked what I was thinking about because I was sitting in the grass staring at the cement pad around a picnic table, (I was watching the ants react to the food dropped while we ate lunch). She thought I was nuts

I think that eventually we will develop more and more complex artificial intelligences, but once we develop one that has the ability to adapt itself (on an individual level, such as an android, or on a social level such as a robotic swarm) then it will be out of our hands. The builder-bots responsible for maintaining and building new robots will develop more specialized robots, and as the specialized functionality is reduced in size the trend will again shift towards generalization to create a more robust system.

The key I think will be promoting systems that can adapt. I also agree that we will never be able to create a functional AI that behaves like a human, however we will be able to simulate human functionality by mirroring all of the chemical reactions that the body goes through each section (there are projects to create this kind of system for individual cells, and I think that the simulated organism is an extension of this).

Share this post


Link to post
Share on other sites
i HATE robots. they are freaky. i was watchin somethin on tv, and this robot in japan has human facial features and can move its face to certain emotions. and one robot could already move by itself and stuff. it was freaky! >.< i don''t want any robot in my house. unless its those toy ones. lol

Share this post


Link to post
Share on other sites
Well, I''m not going to take a solid side on anything only because I don''t think I know enough to do so, but here are my thoughts.

Basically, you can pretty much break down all the complex stuff our brains do into really simple concepts, assuming we know everything there is to the brain. My opinion is that all it will take is research into totally understanding the basics of how the brain works and the processing power to run it.

For example, I''ll use what we kind of know now. Neural Nets, a really sweet concept, seems to "mimic" the brain. It takes the brains complex learning processes, breaks it down to the basics, and achieves complex results. Now, I''m positive theres more to the brain than just that, but for now just imagine that that was IT, that''s all there was to the brain. All you would need is the power to process information like the brain does, and you should get the same result.

It''s of my opinion that all we lack is a *total* understanding of the brain, and the power to simulate it if we do. I mean, its all chemicals and electrical impulses... Once we can break down *all* the brain''s functions to the basics and power that baby up, I''d say hello AI.

Share this post


Link to post
Share on other sites
Here''s an adage I tend to refer to... I think I actually made it up:

If you see gray areas, you aren''t zoomed in far enough to see the small black and white dots.

Dave Mark - President and Lead Designer
Intrinsic Algorithm -
"Reducing the world to mathematical equations!"

Share this post


Link to post
Share on other sites
If someone wants to make AI that advanced but can't afford to build an actual robot, then make a virtual robot in a virtual world. Have a field day.

What you lack in the virtual world is the actual sensors that the robot uses to be aware of everything around it. But you can simulate those, too.

Eyes can be programmed to generate a rendered scene based on the robot's position and orientation, and then shape recognition can be worked into it.

Animation is easy enough, so is collision detection.

Physics and balance routines, etc.

[edited by - Waverider on July 21, 2003 5:59:14 PM]

Share this post


Link to post
Share on other sites
I really don''t think that anyone will ever be able to create an autonomous robot-first, consider the current state of how we create them. Robots are programmed in any number of languages, but the very nature of these languages is a step-by-step, looping process. Basically, a robot or machined is programmed to run through a pre-determined number of steps to accomplish something (anything). We can try to simulate autonomy, but we can never create it-because the languages and the nature of the programming don''t allow it.

The human brain is autonomous. Our mind does not flow in a cyclical pattern, repeating key steps over and over again. Any animal brain (which would have to be essentially reproduced in code or whatever for a robot) is adaptive and beyond our understanding. Robot "brains" are not autonomous, they are simply programmed in such a manner to give the appearance of such.

Basically, the only solution would be to go to the roots-the way we program heresaid "brains." Perhaps we can''t code such things. Maybe the only future of robot AI is through bioengineering and adaptation of already existing animal brains-imagine the possibilities of comprehending a chimp brain and inserting it into a robot body.


"There is no failure without trial; for that matter, there is no success either." - someone

Share this post


Link to post
Share on other sites
Guest Anonymous Poster
to previous post, the brain might not be cyclic but whos to say that with advancing technology such an obstacle could be overcomed. the real problem with robots like in the t3 and other movies is the demand for it.
think about the sony Aibo, it cost thoudsands and it was just a toy, honda has a walking android but its just for research. in the workforce do u realy want to replace a human that cost 5 to 10$ an hour with a (humanoid)robot that cost hundreds of thoudsands plus maintenance, theres no real demand for robots.

in the future there might be with the current interest of the army in autonomos veihcles, because of UAVs like the raptor.

the real obastcle in robotics are physical things like servo motors and powersources

Share this post


Link to post
Share on other sites