Science

Robot Dogs Evolve Unique Personalities

Posted 18 May 2009 at 19:37 UTC by steve Share This

A Phsyorg story details research at the KAIST Robot Intelligence Lab in which scientists use evolutionary processes to develop unique personalities for simulated robots. Their first robot dog, known as Rity, has 14 chromosomes with a total of 1,746 genes that affect the dog's personality and interactions with its environment. The evolutionary processes can go through 3,000 generations every 12 hours, narrowing in on a personality which best fits one of two desired behaviors described as either agreeable or antagonistic. For comparison, researchers used both manual and random processes to generate genomes but neither could match the results achieved through the evolutionary process. The next steps are to improve and generalize the genome structures to allow their use in other types of robots. All the technical details can be found in the paper, Evolutionary Generative Process for an Artificial Creature's Personality (PDF format) from the May issue of IEEE Transactions on System, Man, and Cybernetics.


Changing parameters = evolution?, posted 19 May 2009 at 15:24 UTC by Rog-a-matic » (Master)

I find this experiment interesting but not adequate to convince me that our biological systems advanced to their present state by mutation and selection alone.

In this experiment, the 'chromosomes' are just variables used to define states. I fitness function was used to select one they wanted.

But what we need to see is a system which creates itself by adding new machine parts (fingers, liver, eyeballs), not just changes parameters of an existing machine (hair color). And the rate of change must be faster than the natural forces that work against them (catastrophes and such).

We see many many species going extinct now, but none replacing them. It would be so fun if there was! I'm talking about new animals, not a change in a virus which will never result in what we would need to replace what we're using.

In this experiment it took 3000 generations to change these variables resulting in 2 different personalities. 3000 generations x 25 years = 75,000 years. And this time was needed just to change parameters, not create new machine elements. And from what I can tell, the experiment did not provide random destruction of advances like our universe provides our biological systems.

I believe mutations occur, but they break things, and selection is there to kill off the broken. Just like there are mechanisms in the cells to work against DNA copy errors.

Re "We need" ??, posted 19 May 2009 at 18:17 UTC by steve » (Master)

I think what you mean is that you need a system like that. These researchers are using evolution to develop robots, not trying to convince you that evolution is real. They're probably not overly concerned with whether you "believe in" the science they use to get their job done.

Your idea that mutation can only "break things" is simply incorrect. That's like saying "flipped coins can only be heads" - it's just plain wrong. Not all changes "break things". There are very simple software models of genetic evolution you can use to demonstrate this to yourself if you wish. In fact, I believe we had this whole discussion before...

Yes, we need an example or more faith, posted 19 May 2009 at 19:26 UTC by Rog-a-matic » (Master)

You didn't address my basic points except the one about 'breaks things', but you added 'only' which I didn't say.

Mutations are like rocks hitting an automobile. Within the time of the universe, they will never make it into an airplane. But a rock can break a windshield which could reduce the load of the AC. This may superficially look like an improvement, but it is a removal of information, not adding. It is a breaking of an existing subsystem, not the creation of a new one.

What we need to see, to address these legitimate issues, is to see an existing system add new working subsystems so the net result is positive. That is what is required for an amoeba to become an aardvark. And I am willing to listen to anything you have to say about it and thoughtfully consider it. If you don't have one, I can accept that as an intellectually honest answer.

These are honest questions I have and I hope they are respected and not discounted as taunting.

Re: dull points, posted 19 May 2009 at 20:15 UTC by steve » (Master)

And I think you missed my point which is that the researcher's goal is not to address your disbelief in evolution, it's to build robots. They don't need to take time out from robotics to address your disbelief in evolution any more than they need to address a flat-earther's disbelief in a round planet.

Again, you went through these looney analogies before. Mutation is nothing like rocks hitting automobiles (not even a Mitsubishi Lancer Evolution!), any more than it's like stabbing watches with icepicks, which you claimed in the previous thread. You are perfectly aware of this. These analogies are intentionally designed by creationists to confuse uneducated listeners into misunderstanding basic science (the only alternative I see is believing creationists are incredibly stupid, which, tempting as it is to believe sometimes, I know is not the case).

To be blunt, I think you're being disingenuous. I don't think you're asking "honest questions" or that you're going to "thoughtfully consider" anything except what catechistic argument against evolution to dredge up next.

Flat Earth repells all!, posted 20 May 2009 at 02:27 UTC by Rog-a-matic » (Master)

This is a new record for use of the "flat earth" term. It's only your 2nd posting - high-5!

I didn't pose the questions to rile you up, they are honest questions. I offer an apology for that, and your friendship is important to me. It's OK to disagree, provide a specific rebuttal, or even say you don't want to, but calling my questioning 'looney' is not a very compelling arguing strategy. Sadly, I find this strategy used by both sides too much. I hope you agree.

As you know, it provides me no philosophical crisis whatsoever if evolution is the way things work, I just honestly remain unconvinced at this point.

The story is about evolution so it's valid to talk about what evolution is said to be responsible for in our world.

Not all people who have honest issues with evolution believe the same thing, same for the other side. So your argument that my questions are invalid because of what others think is not very intellectually powerful.

I'll try to direct my writing time to more robots.net stories as you suggested.

You guys at it again?, posted 20 May 2009 at 03:41 UTC by The Swirling Brain » (Master)

Wow! Another evolution vs creation discussion! Wee! Can I jump into the primordial soup with you two? ;-) Just kidding, I really don't want to join in the discussion, I just want to taunt you two a little! Interruption over, go back to what you were doing.

Wait, can I just say one thing. [slaps self] Nevermind.

Helpful info, posted 8 Jun 2009 at 22:17 UTC by steve » (Master)

Experience has taught me it's pointless to argue with creationists. I've also learned I simply don't have the patience for such pointless exercises anymore. At the same time it causes me some distress to see them post misleading psuedo-scientific nonsense. So my compromise will be to post this list of pointers to good sources of accurate information for those who want to know more. If you've read something by a creationist and even briefly entertained the notion that what they're saying is remotely reasonable, you should probably follow a few of these links and brush up your scientific thinking skills.

There is an excellent (and very funny) YouTube series called "Why do People laugh at creationists" that exposes and debunks the misinformation and sloppy thinking of creationists.

http://www.youtube.com/view_play_list?p=AC3481305829426D

Creationists propagate many misconceptions about science in general and evolution in particular, either intentionally or because they honestly don't understand the science. Here's a handy website that indexes most of the common creationist claims and debunks them:

http://www.talkorigins.org/indexcc/list.html

The most common creationist argument posted here is that evolutionary algorithms (and genetic algorithms in particular) do not really work and research produced using evolutionary algorithms must have been "faked" in some way. An assortment of false claims are usually made (e.g., the solution was designed in advanced by the programmer, random changes can only break things or remove information, never improve things or add information). A good place to start is the Genetic Algorithms and Evolutionary Computation essay by Adam Marczyk:

http://www.talkorigins.org/faqs/genalg/genalg.html

A "young earth" creationist may claim the Earth is only a few thousand years old despite all the evidence to the contrary. Here's a good FAQ on how scientists determine the age of the Earth:

http://www.talkorigins.org/faqs/faq-age-of-earth.html

A "geocentrist" creationist may claim the Earth is fixed (does not rotate) and at the "center" of a universe that reolves around the Earth. Here's the quick and dirty science that tells us why they're wrong, followed by a longer critique of geocentrism.

http://www.talkorigins.org/indexcc/CH/CH901.html
http://www.talkorigins.org/faqs/ce/2/part14.html

A "flat earth" creationist may claim the Earth is literaly flat. They generally deny the theory of gravity as well as evolution. Scientists have known the Earth to be spheroidal for thousands of years.

http://en.wikipedia.org/wiki/Spherical_Earth

A few final links that may be helpful on psuedoscience in general, which includes creationism, flat-earthism, young-earthism, geocentrism, and other similar ideologies, superstitions, and conspiracy theories.

See more of the latest robot news!

Recent blogs

15 Aug 2014 shimniok (Journeyer)
8 Aug 2014 mwaibel (Master)
5 Aug 2014 svo (Master)
20 Jul 2014 Flanneltron (Journeyer)
3 Jul 2014 jmhenry (Journeyer)
3 Jul 2014 steve (Master)
2 Jul 2014 Petar.Kormushev (Master)
10 Jun 2014 robotvibes (Master)
10 May 2014 evilrobots (Observer)
2 Mar 2014 wedesoft (Master)
X
Share this page