Slashdot Mirror
Robots 'Evolve' Altruism
sciencehabit writes "Computer simulations of tiny robots with rudimentary nervous systems show that, over hundreds of generations, these virtual machines evolve altruistic behaviors. They begin to share small disks — a stand-in for food — with each other so that their comrades' traits are passed on to the next generation. Experts say the study sheds light on why various animals — from bees to humans — help each other out, even when it hurts their own chances to reproduce."
Does this mean that robots are now more evolved than Randroids?
"Fry: I’m not a robot like you. I don’t like having disks crammed into me unless they’re Oreos, and then only in the mouth."
Yes. Apparently, a few thousand neurons is all that it takes to realize that your own chances of survival go up if you are a member of a group, and that being a member of a group is easier if the other members of the group think you contribute to the group.
Conclusion: Randians have less neurons than bees, and/or a less complex intelligence than these robots.
Those who can, do. Those who can't, sue.
Is there any chance we could get bit.ly and other URL shorteners outright banned from slashdot? Since we're not constrained in character count, their only purpose is to mask the destination of links, which is a bad thing.
Why are the "quotes" around "evolve" rather than "altruism"? The robots did seem to evolve, but what they evolved was tribalism.
Altruism (noun): The principle or practice of unselfish concern for or devotion to the welfare of others - dictionary.reference.com
According to the strict definition, I don't think any theory of evolution could ever explain true altruism, because for altruism in it's pure definition, there simply is no reason. If it has a personal reason, then it is, by definition, not altruism.
Now that's out of the way, there are a number of ways that the less-strict form of altruism (let's call it 'altruistic behavior' rather) would be able to evolve. Firstly, as mentioned in TFA (yes, I skimmed it.. there were only 2 comments at the time) - it makes sense to exhibit altruistic behavior if it improves the odds of your immediate relatives to survive, thereby carrying on part your genes. The more genes your share, the closer the relative, and the more likely you are to care 'selflessly' for them.
But in humans, carrying over genes is not the only reason. There is also the matter of respect, and trustworthiness. In order to convince your allies that you are trustworthy and 'good', you would exhibit selfless acts, with no expectation of return from the person concerned, but definite returns from those you know. By always tipping waiters more than required (selfless by any means), your partner sees your selflessness and gains trust in you. Business partners sees this and are more likely to trust you in business ventures. This all improves your chances of reproduction and survival.
all this is made possible by our fantastic ability to remember and build mental models of specific individuals and relationships, keep tabs on how others acted in the past, and spread the word of any 'egotistic' act to other members of society by means of language. Anyone who is /not/ altruistic (at least as far as others perceives it), is therefore placing himself in distrust, and a disadvantage for carrying over his genes.
So no, it's not much of a surprise that altruistic behavior evolves in robots with a built-in desire to spread their own genes. But it still is pretty damn cool.
altruism and cooperation are investments without guarantee of return on investment. cooperation is not a bartering situation. nor does your effort to redefine trade to be a form of altruism do anything but prove you don't know a fucking thing about what you are talking about
intellectual property law is philosophically incoherent. it is your moral duty to ignore it or sabotage it
I'd like to see the source code and specs, constraints, etc. I've seen robots designed to evolve under certain constraints, that lead to very predictable and obvious traits based on those constraints. For example, if a robot had a goal to pass on its genes, and sharing food was the means to accomplish this, it isn't a surprise that's the result: It didn't evolve that response; it was designed to acheieve it! That's why I'd like to see the actual research. Till then I have to call bogus.
The robots/virtual robots didn't actually evolve altruism as such. I was hoping they were going to say the robots had discovered they ability to recognize weak kin and share food. Instead, the researchers taught the robots how to share, and also changed their optimization problem to "if we both have a decent amount of food, all of our genes will die, but if I give it all away, your genes might propagate." So they just solved the optimization problem they were taught, as opposed to figuring it out on their own.
Their description of the rudimentary nervous systems make the robots sound like they're related to Braitenberg Vehicles, which are otherwise pretty fascinating.
Altruism has a functional definition when referring to evolution, since it's more philosophy if you want to think about animals or bacteria acting morally. I can't recall the precise definition off the top of my head, but it's something along the lines of helping another at personal cost. As I recall, there are three major theories as to why organisms do this.
First is Kin Selection, which is what the article seems fixated upon. Bees and naked mole rats are the classic example. Essentially, it means you'd take a 10% risk of removing yourself from the gene pool to save an individual who shares 15% of your genetic material.
The second is reciprocity. Vampire bats may give a starving individual a blood meal to save their life, and it's a lot more likely if the starving individual offered a blood meal in the past.
The third, and most difficult for people who don't understand math to wrap their head around, is trait group selection. Natural selection has a mathematical model. This is a corollary of that model. In nature, animals form large numbers of groups, either transiently or permanently. Within a group, a non-altruist will always out-compete the altruists and reproduce at a higher relative rate. However, groups with more altruists will reproduce at a greater rate relative to groups with more non-altruists. Overall, you often can have altruists increasing in absolute number despite falling in relative concentration within each group. This process is iterated over generations or within multiple (perhaps infinitesimal) groups that the individual forms within it's life. Being a purely mathematical phenomenal, I would suspect this would emerge within any appropriately complex computer model (it did for the one I wrote for my final project in my Evolution elective back in college).
OTOH, the entire concept of altruism seems offensive to some people. I'm not trying to say any of these are "true altruism", since they happen all the way down to bacteria secreting proteins that deactivate antibiotics, subsequently protecting nearby unrelated bacteria. It's an explanation for observable animal behavior that humans also demonstrate. Plus, "true altruism" isn't a falsifiable hypothesis, so there's little sense in arguing about the moral proclivities of humans, bacteria, chemicals, cultures, or ideas.
The genetic reward is proportionate to how much of one's genes the recipient shares. Thus altriustic behavior will (and should) drop off outside of children, of family, of relatives, of tribe, finally of all of humanity... however, it never reaches zero as long as the recipient is approximately inside our species.
And there is the confounding variable that because society rewards altriusm (for obvious reasons), individuals will invest in appearing to be altruistic, especially if they actually are not altruistic. Such behavior will overwhelm the very mild altruism that we are looking to observe between strangers. You need to track down some of those "subject is not aware he is being observed" experiments.
FATMOUSE + YOU = FATMOUSE