Is Compassion a "continuous pattern-sympathy"?

My criticisms of Ben Goertzel's: Universal Ethics : The Foundations of Compassion in Pattern Dynamics
URL for above: http://www.goertzel.org/papers/UniversalEthics.htm

by David Clark Oct 27, 2004

• How can a pattern "tends to continue into the future". By what mechanism or criteria do they "tend" to survive.
• Survival of the fittest?
• Random chance?
• Who's your daddy?
• Patterns are created by some process or algorithm. Looking at the pretty pictures might be nice but how does that create meaning when the patterns are the "result" not the source.
• Even when a pattern is created from a source pattern (as in the Life game), it is the algorithm that creates the new pattern from the old pattern not the old pattern itself.
• You say that if "continuance over the long-term is reasonably well-predicted by continuance in the short-term" then there is "continuous pattern-sympathy".
• What criteria determines "short-term continuance". Is this criteria not the predicate for the continued existence of the pattern rather than the pattern itself?
• Ray Kurzweil says that there will be a singularity by the year 2025-2030 by showing the exponential growth in many things over the past number of years but he doesn't say how it will be made to happen.
• He certainly doesn't say that he will be part of making it happen.
• He doesn't even give the motivation for why people are likely to fulfill his extrapolation.
• Talking about the end result as if it had it's own volition, without regard for how you are to get there, doesn't make much logical sense to me.
• The word "sympathy" I understand. The word "pattern", I understand. I (a human) can sympathize with something or someone. How can a pattern (a non living entity with no physical presence) sympathize with anything?
• You say "pattern in each one of the minds "wants" to perpetuate itself".
• What does it mean for a pattern to "want"? Is this not the same as saying my fork wants to pickup food and put it in my mouth?
• I could design an algorithm that could have a goal that could be anthropomorphized as a "want" but a pattern is the leaf, "the end result", "the picture that is the result of".
• You equate "compassion" to "patterns in my mind are compassionate toward their clone or near-clone" but people are compassionate all the time with people who are very different from themselves.
• Many people have compassion for others that think, act, and have a totally different culture than themselves.
• Could it be that some people have a desire to see themselves as "good"?
• Why couldn't being compassionate towards others just be a manifestation of proving to yourself that you are "good".
• Many people will say that their "compassion" for someone else was fulfilling even if the net result of their compassion was zero for the other person.
• I think compassion has nothing to do with the person who is the recipient of the compassion and totally about the person who wants to feel it.
• You say that there are other reasons for compassion (like familiarity) but I have seen compassion many times where the person giving has no apparent connection to the recipient at all.
• Many people are very compassionate to animals that are not even theirs. How do explain that in terms of similar patterns?
• Does the frequency of a pattern determine it's value?
• If you make 10 weak arguments and I make 1 really weighty argument, do your arguments automatically win because you made more of them?
• When diamonds are mined, which is there more of, the kimberlite or the diamonds?
• If you give value to something because there is more of it, would you throw out the diamonds and keep the kimberlite?
• You say "natural selection, at bottom, is nothing other than a particular instance of the general phenomenon of continuous pattern-sympathy.".
• Whether a pattern actually makes a better "fit" in it's environment has nothing to do with "natural selection"?
• I guess there is a "general phenomenon" that the winner in a "to the death" fight might get to produce more offspring and the loser definitely doesn't. Is this what you mean by "continuous pattern-sympathy"?
• Natural selection is the idea that if something gives an advantage to an organism that results in it's greater genetic propagation then in the long run that organism will persist in that ecological niche where it's competitor might not..
• If you want to call that change (advantage) a pattern that's ok but it is totally contextual. The same pattern might fail hopelessly 10 feet to the right. The value of the pattern cannot be extricated from it's context. It also might lose tomorrow if a better pattern comes into this context and out competes the first pattern. The pattern by itself has no special value.
• You say "long-term survival of the short-term survivors," which is totally obvious but misleading.
• Your can't have long term survival without short-term survival but the reserve is definitely not true.
• If it was, then no species would ever go extinct.
• If we had species A and B that are now alive. You would have to say that in the short term (the next 1 minute or 1 second etc) they would both still be alive.
• Given that you say that the short-term continuance is a predictor of the long-term continuance then it would follow that both A and B would exist in the long run.
• Genes don't "want" to propagate themselves into the future.
• Genes or "patterns of genes" or "groups of genes" have no volition.
• Their being propagated into the future is a consequence of them winning the fight.
• They don't even "know" who or what they are fighting. The "fight" comes down to the fact that if organism A acquires the needed resources of organism B then they are simply unavailable to sustain organism B.
• You say "phenotypic pattern that has persisted itself over time".
• You anthropomorphize a pattern as if it was a person.
• The successful "phenotype pattern" is the consequence of a successful battle.
• But the battle was not waged by the pattern, rather it was waged by the process of natural selection which derives from the fact that real matter can't exist in two organisms at the same time.
• You wrote "Compassion itself, as a trait of organisms, is a particular pattern that spreads exponentially according to the logic of continuous pattern-sympathy. And it spreads faster than would be predicted via looking at gene-level dynamics alone."
• How can a concept like "compassion" which is an idea made up in the minds of humans have anything to do with our genes? Do you have any scientific data to show that "any idea" people invent has anything to do with (is derived from) their genetics? I have seen no such data myself!
• I believe myself to be compassionate but I don't see any exponential spread of compassion. If it's spread was exponential then all persons on Earth would already by compassionate and I don't see anything like that out there.
• Compassion is seen most often in people who are not just concerned with their own well being. Most people are definitely NOT like that.
• You say "[compassion] emerges naturally from natural selection" but I see exactly the opposite.
• Natural selection is most often not compassionate at all. When it looks that way on the outside, we investigate that odd occurrence because we don't expect it at all.
• We normally find a reason for the "apparent" compassion which shows why the "pattern" is beneficial to both parties. For the one's we haven't found the reason, we just haven't found it yet.

Conclusion:


I think gardeners think of the world in terms of plants and how they feed people etc. They would probably define "compassion" using the gardening terms that they know. Ben Goertzel, IMHO, has defined evolution and "compassion" in terms of his mistaken belief in a "swirl of patterns" creating intelligence all by itself. I think he is wrong about compassion and wrong about "dynamic patterns" or "emergent patterns" creating any useful intelligence.

Some would argue that our existence is proof that intelligence can be created from non-intelligence by this method. Our brains are a grouping of many specialized and general purpose neural nets. We have evolved over the last 2 billion years with unimaginable physical contexts that have molded us into what we are. Is it possible to build a neural net (large enough) and duplicate this environment (all the diverse environments of humans and all of the environments of our pre human ancestors) in any reasonable time scale? The bigger question is "Is this the quickest way to get to intelligence created with today's technology?".

I believe that it might be possible to create an artificial brain by duplicating the neural network of the brain and then teaching and nurturing that AI for next 30-100 years. Even if we could build the machine physically, we don't know exactly how to teach it as we teach our own kids. The end result however, would be hardly better than a human's brain and full of the flaws that plague human thinking. (Lack of focus, faulty memory, slow cycle time, etc) Why would such a machine have an easier time increasing it's intelligence as compared to a human when it would have inherited our bad as well as good thinking memes? We already have 6 billion brains on the Earth so why create a brain that is no better than what we already have lots of?

Neural nets have their place but even when they work, you never get the reason or the logic flow of why! This is not a problem if the neural net is used for pattern extraction in video images, sound extraction, character and hand writing recognition etc. All that matters is that we get the end result, not how we get there. This is not true for most other kinds of problems.

Evolutionary algorithms have a place but creating appropriate success functions for each generation has only limited usefulness in some class of problems.

Agents that can interact with each other also have a useful purpose. If I was running a simulation of a group of people and I wanted to model the interaction of that group on a subset of human properties, I might create that model using agents. However, most other classes of problems don't lend themselves to the agent metaphor.

The path to intelligence will not happen on it's own. It will only happen if we make it happen. No amount of processing regardless of how smart the algorithms are, will unlock the secret of intelligence without injecting vast amounts of information and using a multitude of different techniques. This has been proven (at least to some degree) by the failure of all the AI projects to date.

It could be that some form of "evolutionary" programming will help with creativity. However, humans rarely need to be creative to be considered "smart".

I feel the best way forward is to create a system that makes models and abstracts those models so that algorithmic solutions that are successful in one context can be experimentally tested and/or used automatically in other contexts. These algorithms would have to be programmed by a human for the first long while but most of us don't create totally original ideas either. Most of our original thinking is using an existing idea in a new context or using an original group of existing ideas in a new context. This could be done for an AI. The idea of making an AI that has information grounded in reality is a good thing. However, if the intelligence of a person was limited to only what we could acquire with our own two hands, we would not be very smart at all. The idea that you could have a "human plus" intelligence without a massive database of real world information doesn't hold up to what we see in people. We don't believe anyone is very smart unless they can hold a meaningful conversation that entails much existential information. I don't think that just a huge amount of information by itself will make the AI but I can't see how any AI could be made without it.

The core of the AI must be able to accommodate many different kinds of techniques and integrate them into a whole so that they can be useful. One way just won't work.