jump to navigation

Robot Cheaters and Heroes February 12, 2008

Posted by Nina Rosenstand in Artificial Intelligence, Ethics, Nina Rosenstand's Posts, Science.
trackback

It took 50 generations of robots evolving from basic light-sensitive wheeled mechanisms to something much more sophisticated—but now the Laboratory of Intelligent Systems at the Swiss Federal Institute of Technology can boast of having created four groups of robots who have evolved into light-consuming and communicating entities. Three out of the four groups will alert the other robots when they “find food.” The final group has developed robots who will lie about the food source, telling the others that it is poison, and then eat it all themselves. And if that isn’t enough, some robots have evolved into heroes who will alert others to danger and die saving the others. In The God Delusion Richard Dawkins (The Selfish Gene) calls human altruism a “precious Darwinian mistake.” So does this mean that any evolving, communicating entity will travel along the same path as we humans have? Here is a quote from the original report summary:

 “We conducted repeated trials of experimental evolution with robots that could produce visual signals to provide information on food location. We found that communication readily evolves when colonies consist of genetically similar individuals and when selection acts at the colony level. We identified several distinct communication systems that differed in their efficiency…. Under individual selection, the ability to produce visual signals resulted in the evolution of deceptive communication strategies in colonies of unrelated robots and a concomitant decrease in colony performance. This study generates predictions about the evolutionary conditions conducive to the emergence of communication and provides guidelines for designing artificial evolutionary systems displaying spontaneous communication.” 

So the “liars” were unrelated to the others, while communication went smoothly if the individuals were genetically similar. Without having read the entire report I will jump to the conclusion that the “heroes” came from the genetically similar groups. But does this prove that Dawkins is right (Moriae, weigh in!), or that this is no “mistake” at all—that self-sacrifice will happen, because being a member of a colony fosters genuine selflessness? Then again, maybe the researchers at the Swiss lab have merely reinvented an ant hill…

Advertisements

Comments»

1. Stephen - February 12, 2008

The problem is that programs have a finite amount of choice, and “genetically similar” programs will have little deviation in their range of choices in relation to one another. Because of this, it is logical that differing programs (the liars), would react in a different manner.

While it can be argued that humans, like these robots, are all “programmed” differently, human beings generally make altruistic decisions based on emotion or instinct, which cannot be programmed into a robot. Therefore, Dawkins belief does not really seem to apply, as AI cannot explain human behavior (See Searle’s Chinese Room).

2. Moriae - February 15, 2008

I’m wondering if some people who haven’t read Dawkins might misunderstand the subtle issues being raised here. He (Dawkins) doesn’t mean ‘mistake’ in the way many may assume. Let me quote Dawkins (p. 220) to draw out this point:

“It is important not to mis-state the reach of natural selection. Selection does not favor the evolution of cognitive awareness of what is good for your genes. That awareness had to wait for the twentieth century to reach a cognitive level, and even now full understanding is confined to a minority of scientific specialists. What natural selection favors are rules of thumb, which work in practice to promote the genes that built them. Rules of thumb, by their nature, sometimes misfire.”

At this point he offers the example of the famous ‘mistake’ of the reed warbler birds who often unwittingly raise the offspring of cuckoos. They do so simply because their instinct is to feed whatever jabbering birds are in their nests. If you replace their own birds with others (as cuckoos do), the instinct continues unchanged and they end up raising another species of a bird in their nest uncomprehendingly.

Dawkins continues (221), “I must rush to add that ‘misfiring’ is intended in a strictly Darwinian sense. It carries no suggestion of a pejorative.”

“The ‘mistake’ or ‘by-product’ idea, which I am espousing, works like this. Natural selection, in ancestral times when we lived in small bands like baboons, programmed into our brains altruistic urges, alongside sexual urges, hunger urges, xenophobic urges, and so on. An intelligent couple can read their Darwin and know that the ultimate reason for their sexual urges is procreation. They know that the woman cannot conceive because she is on the pill. Yet they find that their sexual desire is in no way diminished by the knowledge. Sexual desire is sexual desire and its force, in an individual’s psychology, is independent of the ultimate Darwinian pressure that drove it. It is a strong urge which exits independently of its ultimate rationale.”

I’ll offer two of my own examples at this point to illuminate how such ‘misfiring,’ ‘mistakes,’ and ‘by-products’ (my preferred term) are used and manipulated for good or ill.

It’s a well-understood point of evolutionary psychology that one’s genetic interests are served by practicing ‘altruism’ with genetically related members in a family. The closer the ties (son or daughter) the greater this sense will be, and conversely, the greater the genetic distance (say, a third cousin) the more this sense will be absent. Yet we can easily get this sense to ‘misfire’ for whatever purposes we may have in mind.

Do you think it is incidental that in the cases of the Marines and the Hell’s Angels they commonly refer to each other as ‘brothers’ (“band of brothers”) and ‘family’ when in fact no genetic link is ever present? But it is a useful ploy when the existence and cohesion of a particular group hinges on a sense of connection that would be absent if viewed from a purely rational perspective. So the fact that people use such ploys suggests how genetically disarming it is to use appeals to “family” when trying to cajole others for another purpose (notice how in politics and religion this is prominent too?). Dawkins mentions this kind of connective ‘misfire’ in our caring of abandoned children and/or our feelings for an adoptee (p. 221).

My second example is curious, but it also suggests a noticeable ‘misfire’ that is often ignored. Consider the genetic connections between a husband and wife. Unless they are from the Appalachian mountains or from Iraq (cousins marry there constantly to maintain tribal cohesion), the likelihood of a genetic connection between the two (in all other cases) is missing. But in their early years of raising a family a kind of truce is maintained because the two have a shared interest in raising their children, despite the fact that each child only has half of each parent’s genes.

Yet imagine later conditions with grown children. The ‘truce’ has certainly played-out its usefulness, and yet two totally unrelated people get along (most of the time) under one roof. Now notice the not too uncommon phenomenon of an elderly husband defending his wife to the death. Was it gratitude? That seems a very unlikely motive. This suggests another ‘by-product’ (misfiring) being displayed where through time and general utility a sense of familial bond was established when in fact the interests being served in losing your life for your wife are genetically useless. In such cases we generally use the rubric of ‘love’ as the most useful explanation for the behavior.

3. Nina Rosenstand - February 19, 2008

Interestingly, in The God Delusion Dawkins calls this phenomenon a “blessed Darwinian mistake.” So you’re correct, Dawkins doesn’t mean that we are stumbling bunglers with no idea who are our relatives and who aren’t, and he is appropriately appreciative of the phenomenon of altruism. But if we try to explain all human behavior as “selfish,” we end up with an explanation that is too broad. Some forms of “selfish” will then have to be classified as “more selfish,” and “less selfish,” as Mary Midgley points out. And, as she also points out, sometimes we elect to behave unselfishly (‘scuse me, less selfishly) knowing full well that the recipient is a stranger. Then whose mistake is it? “Nature’s”? A little too anthropomorphic/poetic for me. The newer analyses involving a brain center for empathy (which can be overridden by logic/selfishness) will probably turn out to have a much higher explanatory success than the selfish gene theory. It would provide a better explanation of your example with the husband dying to save his wife than “selfishness misfiring.” But I’ll bet the debate won’t go away! Because it’s fun!

4. Huan - February 20, 2008

I think the argument about the varying degrees of selfishness just kind of trashes the word completely. The word was created under false assumptions about human psychology to highlight its brother selflessness I would assume. Now days these words are doomed to suffer technical inaccuracies, just showing how often we wrongly categorize concepts based on false beliefs. (maybe we need a new word or something.) Well more importantly I don’t think the whole selflessness is utterly impossible argument is some how a smart-ass comment on the technical mistake of the term, it actually points out a fundamental aspect of human psychology and shows how dogmatic altruistic morality causes too much conflict within the human psyche and becomes either poorly practiced or blindly followed.

I totally agree that the concept of a selfless gene is a little too much of a stretch, I will even go further to say that the empathy brain function is not necessarily meant for empathy at all. It seems to me that the emotional responses of the brain are merely tools at our disposal that has varying functions depending on the context of our perception. Using the example of music, it could easily evoke the same emotional responses as empathy or that of moral judgment physically speaking, yet completely lacks any context. This shows that empathy and related emotions are more context based than physically based, perhaps leaving some room for a Why. Why does watching a puppy get tortured really evokes my emotions? The answer is perhaps far far more complex than it’s just how our brain is programmed.

Relating this back to the original topic, I don’t think theres some kind of genetic misfiring at all, since the concept of selflessness is more nurtured based than anything. Wouldn’t it be weird to assume that the turning of a screw is done by the screwdriver and not the force acting upon the screw driver?

5. Moriae - February 21, 2008

I was having a hard time figuring out what you were driving at, even if it hadn’t been with a screw driver. The issue has to do with Dawkins’ concept of a “selfish gene,” not an ‘unselfish gene’ as you often wrote. Is this perhaps a case of a kind of Roseanne Roseannadanna where we’ll hear a “never mind?”

He admits in his recent release of a new preface to his book, The Selfish Gene, that the whole imbroglio regarding his use of the term ‘selfish’ has been unnescessarily distracting to the real thrust of his argument. I would think that before we waste any more time on this it would be profitable for us to familiarize ourselves with this writing first. If not, this could get unnecessarily out-of-hand.

Let’s clarify issues before we take sides. Do we really think we change minds here? It’s unfortunate that much too often people parse an issue to such a partisan degree that chances to inform become instances of misinformation. Obfuscation merely makes us have to write more here to clear-out misunderstanding, and I don’t type well!

6. Huan - February 21, 2008

I was under the assumption that what you called a misfire was about how heroic or empathetic reactions are due to a “bonus feature” or “side effect” from the selfish gene, right?

Now at some point I lost track of the origin of this, and selfless/selfish became the same thing, since the theory implies that they are different effects of the same gene. It is kind of funny how I started talking about the selfless gene though, didn’t really notice that.

Anyways, the question I was talking about was whether selflessness was some kind of genetic side effect of the selfish gene. Responding to Professor Rosenstand’s reference to the empathetic brain center, I attempted to point out that all this talk of specific genes and specific functions they serve is extremely slanted, reminds me of when scientists explained all animal actions with “that’s just their instinct”. Going back to the last article about morality and empathetic, judgmental emotional responses, it referred to a portion of the brain that activates when someone reacts emotionally to something. I’m not sure whether this is what Professor Rosenstand was referring to, but that’s certainly what it reminded me of. I suppose my point was that these portions of our brains or even genes are not the sole cause of what we label them as, the selfish gene, the selfless gene, the empathetic brain portion, etc. They’re just tools at our disposal, and the actual emotion invoked is extremely dependent on context of what is perceived, whether it’s selflessness or selfishness.

I suppose it’s just a different perspective on the selfish gene theory. Instead of the selfish gene being the cause of some side effect that creates selflessness, they’re both the result from what I see as just a tool, a screwdriver, the screws being loosened(selfless) or tightened(selfish) depends on the forces acting upon the screwdriver. (perception?)
Sorry for the metaphor perhaps getting out of hand, got a little cheesy.

7. Huan - February 22, 2008

Oh and I totally see what you mean by the partisan thing leading arguments nowhere. I’m not really trying to take sides, just trying to add my perspectives on the topic.

8. Moriae - February 23, 2008

Cool. Anyone who writes here should be considered friends first, and only after you get ticked off should you hunt them down. 🙂

The AI thing I find very humorous. I bet most of the people working on this stuff had a hard time getting dates in high school. I wouldn’t underestimate a woman’s intuition even in high school. My suspicion has always been that what passes for intelligence is merely a mitigated example of what we call an idiot savant. It’s just in the cases of what passes for ‘smart’ is much less capable that what some of these special savants can do, and what these idiot savants are deficient in is almost matched by our ‘brights.’ If we expect ‘intelligence’ to be noticed by these kinds of people, God help us. Most are probably split on Obama or Clinton— I REST MY CASE.

9. Huan - February 24, 2008

Yeah I wish they went in more depth on how exactly these robots were set up, it’s quite likely they’re a bogus imitation of intelligence and sort of a self-fulfilling experiment. I think we should probably spend more time on reverse engineering the brain. 🙂

10. Moriae - February 24, 2008

I wonder: if you merely “heard” the voices of Obama, Clinton, or McCain coming out of a box, would you really suspect some kind of ‘intelligence’ was the cause. And if we disconnect the unwarranted conflation of “human” with “intelligence” (notice how susceptible we are to flattering notions–it will be the death of us I’m sure), what model of intelligence would we use?

It would be enough to confuse Searle I’d think. And if we couldn’t convince Searle that signs of real intelligence emanate from a box despite the fact that they pronounce the very words that wow our contemporary rabble, why should we respect the people who really believe these broken-record mannequins running for office are convincing specimens of intellect?

Or consider: if people consider others ‘intelligent’ and/or smart simply because they are smarter than they are, in all truth, how smart do these politicians really have to be? Anything seems “above” bottom dwellers.

So if millions of Americans (ones who would, by the way, seriously challenge the notion that they may be in some way wanting themselves when it comes to intelligence–desipte the ample evidence) cannot detect signs of “artificial intelligence” in political candidates that warm the cockles of their hearts, how would we convince the American canaille what AI is like when scientists claim to have achieved “real” AI? Would the AI creations have to include compelling likenesses of Obama, Clinton, or McCain to pull off the trick?

That’s why I don’t believe we’ll EVER know about extraterrestrials ‘smarter’ than us either. If they aren’t as smart, they couldn’t get here; and if they were smarter than us, then they wouldn’t come here. They couldn’t be too smart if they visit New Mexico.

Besides, they likely wouldn’t want to know us for the same reasons you choose not to listen to your next door neighbor. And being smarter than us they’ll also have more and better reasons for avoiding us—which will be easy for them, because there are so many places for them to hide from us.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: