The problem is that he is assigning goals to cellular activity. Very Aristotlean. I doubt that your fingernails are particularly interested in protecting the soft skin underneath. I am utterly unconvinced that they do it because they thought it through and concluded that they ought do so. If we had 28 different pathways for fingernails to grow depending on conditions (even if one involved application of electrical current), it wouldn't make me more convinced that my fingernails were exhibiting intelligence.

I've no particular problem with someone thinking about the world in terms of objects having inherent goals/purposes other than we kind of ran into some major dead ends in learning about the world when we used to think about it in such terms. Perhaps there is something useful in such ways of thinking, but I am not smart enough to determine what they might be.

fingernails that incorporate fitness of fingers into their own fitness may well be fitter than ones that dont.
and fingers that incorporate fitness of fingernails into their own fitness may well be fitter than ones that dont.

so evolved 'symbiotic' stratagies may be fitter. Leading to fingernails that adapt faster to changing finger needs than we might otherwise expect.

This can happen when over very long periods finger fitness changes at times and those with less adept fingernails die out

Still dumb as **** but it can look real clever.

If a few nearby distinct fit solutions exist then evolving to 'quickly' switch between then can be very fit and look like magic. It's also pretty easy.

Levin talks about this. My understanding of what he said is that we've gone overboard in shying away from this kind of language. If it's backed up by a standard of objective empirical specifications then it can be useful in driving empirical research. However, he later looks at hypothetical entities operating on who knows what kind of physical systems, e.g. AGI, BIO-AGI combos etc., which rate extremely high on these measurement standards for intelligence. He argues that we may need to adjust our ethical approach to such entities. This seems a little bait-and-switchy to me.


PairTheBoard

My fingernails are very dirty at present.

It is bait and switch. Some people (for some weird reason*) think that intelligent things deserve to be treated in some sort of moral way. If you change what you mean by "intelligence" you can't legitimately carry all of the connotations that went along with the original definition to the new one.

I have some really nice excel spreadsheets that are pretty smart. They even adapt to new data, putting out new output. They never cry when I turn off my computer.

What is "the original definition" of "intelligence"? I think that's Levin's complaint. Armchair philosophers can sit back and talk about it but can't provide a definition that can be studied empirically. He also wants a definition that provides for a non-binary sliding scale that accommodates a continuum of intelligence. So he goes with the definition proposed by William James.

He notes the difficulty with this philosophically and intuitively. At the smallest scale people will say, "this is just physics". At the largest scale they will say, "this is just emergent complex neural activity". He claims it's all the same and just a matter of scale.


PairTheBoard

I meant, by "the original" as the one used to justify a need for ethical considerations for a being. This would still be silly as we now mostly believe that it is the ability to suffer that would justify ethical treatment.

"An organism that demonstrates the ability to learn from experience" would be one such scientific operational definition of "intelligent being" that can be empirically studied that he apparently doesn't think exists. There are other operational definitions that also exist and can be empirically studied, such as the ability to write a sentence.

That is beside the point. You can cut up a planarian into bits. They have it on record on how much it can be cut up and yet grow back to a full planarian.

Supposedly Levins experiment has an altogether different conclusion. In that you can train a planarian. What train means with regards to a planarian is not quite clear to me. But itÂ’s something along the lines of exposing it to certain external stimuli that makes it act a certain way.

But then you chop its head off. It doesnÂ’t merely grow back but it grows back still acting as itÂ’s trained. We shouldnÂ’t be glossing over this but trying to understand it. This question is why Bernardo is interested and what levin canÂ’t answer

The Philosophy of Intelligence is discussed in this interview with Levin by "Cognitive Revolution". The first 25 minutes is mostly review of results. From 25 minutes on they get into the philosophy. Comparisons with AI. Need to support the philosophy of intelligence with experiments to see what happens when you perturb the system. Why there's a payoff treating some things rather than others with the tools of behavioral science. e.g. Training animals. What things scale up. The phenomenon of unexpected capabilities. Delayed gratification - going away from a goal to get closer later. Learning from experience.

Looking at concepts like Machine - Life - Human - Emergence, as engineering protocols rather than objective truths. Utility of certain world views as seen from another perspective. e.g. You want an orthopedic surgeon, but not your psychotherapist, to view you as a machine. The element of "surprise" in emergence depends on your perspective. Scaling down to the smallest.




Dr. Michael Levin on Embodied Minds and Cognitive Agents
Cognitive Revolution




PairTheBoard