This episode really got me thinking. I had to pause the audio when the issue came up of integrating neurodiversity into the discussion of “possible mind space” in artificial general intelligence.
I can't claim credit for the "possible mind space" concept; I believe I first encountered it in the context of a talk by Eliezer Yudkowsky -- it is an extremely useful concept, though, and one that is important to keep in mind whenever postulation about what the term "mind" means in the first place. One trap that must be avoided with regard to thinking about cognitive enhancement, artificial intelligence, and the "medicalization" (or lack thereof) of existing neurophysiological variations in sentient creatures is that of working under the (possibly unconscious) assumption that the range of possible self-aware processing mechanisms that can be thought of as "minds" is far narrower than it actually is.
While recognizing that each existing person is quite unique even in the midst of an assumed-normative range of neurological and biological characteristics, it does seem to be true that most people hold some personal notion regarding the nature of a "normal" mind, and that most people do have difficulty recognizing the potential validity of minds that differ sufficiently from their own. Some of this is probably rooted in the lack of a coherent definition of "mind" in the first place (or at least, a widely-accepted, concise, parametric definition) -- that is, a person might avoid classifying a given piece of phenomena they encounter as a "mind" not due to any kind of prejudice or lack of imagination, but due to the fact that they themselves simply don't know what sorts of phenomenological data point to a mind "being there" at all.
The question, from the perspective of someone encountering a phenomenon that could potentially fall into the category of "mind", commonly ends up being a rather utilitarian one: if a mind exists in one's vicinity, different obligations are owed to that mind than would be owed to an inanimate object (though it should be recognized that there could very well be minds in the "potential mind-space" that would actually prefer to be treated as one might treat a toaster -- we simply don't know!). A mind-possesing entity will, conceivably have some set of native demands; at the very least, it will require resources -- material, energy, informational, what have you.
So, when the "known" mind (say, a present-day human person) encounters a given piece of unfamiliar phenomena, the question of whether this phenomenon has a mind, or is a mind, becomes pertinent, particularly when the human is faced with issues of resource scarcity (perceived or actual) or information need (since the identification of a particular thing as a mind could end up being tremendously important if, say, the human is in the midst of attempting to solve a problem that the local set of available minds has found too perplexing to make much headway on).
Humans, at present, do not have much in the way of sophisticated means of (a) detecting minds, or (b) establishing the validity of an identified mind. Humans tend to rely not only on behavioral signals, but on typical behavioral signals -- many of which can be culture-typical.
For instance, an American who makes little eye contact might be considered "shifty" or untrustworthy -- or possibly autistic, and many nonautistic humans have difficulty acknowledging the validity of autistic cognition. However, a person who makes less eye contact is born into a culture where direct eye contact is considered rude or presumptuous would not stand out, even if his or her instinctual predilection is not to make eye contact based on neurological factors.
If the variation in behavior is limited in this example to that of eye contact (or lack thereof), it seems quite clear that while a mind might be recognized in both cultures as existing, that mind will only be considered fully valid in the country where direct eye contact is not the social norm.
When you add in other behavioral variables -- responsiveness (or lack thereof) to verbalizations from other persons in the vicinity, presence or absence of typically-communicative speech, apparent ability to carry out what are considered "basic" life skill functions -- you end up with a very complex set of external factors that are commonly used in the assessment of a person's inner life, even when they may have little to no correlation with how the person actually perceives and experiences reality. It amazes me that people who have no problem believing that they'd be able to recognize a manifestation of artificial intelligence could have trouble accepting the idea of a person who does not speak or look anyone in the eye, but who is also fully capable of experiencing a complex inner life.
If AI were to emerge "inside a computer", would this AI only be recognized as valid if it created an avatar that looked human, and that provided culture-typical human responses (perhaps in the context of a two-way conversation, in which the avatar on the screen interacted with the programmer as one might with someone on a video-conferencing system)? Or, is it critical (as I believe it is) for humans to avoid narrowing the scope of "potential mind-space", perhaps to the exclusion of a myriad of valid minds, some of which might communicate in ways that would not be recognized as typical?
One issue I find quite intriguing is the question of whether the existence of a mind necessarily means that that mind is capable of communication with other minds -- and if so, how can we work on sorting deliberate communication from random (or perhaps, more properly, pseudorandom) phenomena?
A less behaviorally-based means of determining the presence and nature of minds is imperative -- or at least, the spectrum of what is to be considered purposeful and/or meaningful behavior must be widened considerably. This doesn't mean assuming everything anything does is purposeful or indicative of a mind (or a valid mind), but rather, seeking out and acknowledging "proof of concept" examples of cases in which existent minds do not conform to the expected behavioral manifestation that would typically be expected to correspond with a mind of particular complexity.
5 comments:
You have touched on a subject which has occupied my "mind" for quite a while. Assuming the emergence of some new kind of intelligence, one hopefully orders of magnitude more capable than ours, how would we recognize (and hopefully interact) with it? I agree with you that the process could not be based on behavior as that would be too subjective and narrowly applicable to human-like intelligence, which is only one example from the set of possibilities. I like to think that any form of intelligence would have the ability to create, rationally evaluate, and logically pursue goals. But how could we become aware of something like that happening? Will we just someday realize that there seems to be a lot more serendipity in the world, that a lot of problems just seem to be resolving themselves? I don't know. I often wonder if the new mind, with all its intelligence, will deem our awareness of its actions either unnecessary or undesirable to the pursuit of its goals, whatever they may be. In any event, I'm intrigued by your suggested "proof of concept" approach and wonder if you could elaborate further.
I'd say that whether or not a mind is capable of communicating with other minds depends mainly on the surrounding environment, rather than the nature of the mind itself. Communication always requires a functional interface, but the existence of a mind doesn't.
If you could keep a brain alive in a jar after the body died, for example, how would you know what it was thinking?
There have been some creepy experiments recently with building neural networks out of rat brain cells in a petri dish. Do these cells have a mind? How would we know?
ekiech said:
I'm intrigued by your suggested "proof of concept" approach and wonder if you could elaborate further.
Well, we don't have artificial minds (that we know of) yet, so the proof-of-concept space we have available consists of that comprised of humans and other animals.
Some "proof of concept" examples have already prompted scientific and cultural paradigm shifts; for instance, the identification of particular kinds of intellect in animals such as Great Apes, dolphins, and elephants.
It only takes one sign-language-using chimpanzee to open people's minds to the possibility that "human" isn't the only descriptor that properly goes in front of "mind", even when you're only counting presently-existent, nonartificial creatures.
Of course, judging the presence of a "mind" by noting an ability to communicate in a human-understandable fashion is very primitive and crude and I hope we move beyond that someday through some means, but at the very least, meaningful cross-species communication of any sort proves that there are a lot of human-borne assumptions people make (with regard to what the "necessary precursors" are to minds and intelligence) that can cause the existence of other kinds of minds to remain hidden from us until we shift our perspective appropriately.
Once the mind is revealed it can seem obvious, but it's important that we never forget how invisible that mind seemed before we looked for it in the proper way.
Again, this doesn't mean we need to assume every piece of phenomena we encounter has a mind (or is a mind), but rather, that we avoid "human racism" in the assessment of potential other minds.
And there's also the matter of the existence of *humans* that defy attempts to assign the assumption of "mind" based on narrow behavioral and developmental skill-set notions.
Many humans tend to think in a manner which overlays all kinds of assumptions onto another person's development -- for instance, parents will read a child-development book and assume that if their kid hasn't learned Skill A, they're never going to learn Skill B, when in fact Skill A and Skill B can probably be learned completely independently of one another.
Hence, a child could very well develop Skill B, only the parents wouldn't even notice because they are too busy watching for Skill A, which might never appear or which might appear much later. People need to stop assuming that, by default, some skills are inextricably "connected" to other skills -- if not, they can miss evidence of "mind" that is actually quite blatant.
abfh:
First of all, glad you found your way over here (I thought we might be navigating in some of the same circles, and I'm pleased to see that I am not the only one who understands that the future needs all kinds of brains.).
As for your "brain in a jar" experiment -- I've often wondered about that, and one thing that I am curious about is whether a mind must have some means of getting environmental feedback (and processing it) in order to experience consciousness.
For instance, if a creature were born unable to see, hear, smell, taste, or feel (in the temperature or tactile sense) anything, could it actually "think"?
In humans, we know that the phenomenon of dreaming can produce a form of consciousness that happens entirely in our minds (with only minimal "interference" from external stimuli; we sometimes incorporate the ringing of an alarm into a dream, but not always), but could those dreams retain their character if the dreaming organism had never experienced any kind of sensory input whatsoever?
I don't think there's any conclusive answer to that question based on what we know at present, but I do agree with you that communication facility is quite dependent upon the environment (and also, perhaps, what technologies are available in that environment).
As for whether networks of rat-neurons have a mind, again, I have no idea how we'd discern that. It would seem that the mind-detection problem is somewhat based on human limitations -- though how we'd expand or enhance our own abilities to be able to recognize minds remains to be seen.
I think people can at least make a start at this by moving further away from basing estimates of mind on visible outward behaviors that fit a particular pattern that the observer is accustomed to associating with the presence of intelligence.
Of course, at this point, we are still limited to making observations of things that human brains can recognize -- but we can certainly work on expanding the range of what is recognized as purposeful and/or communicative.
This doesn't mean anthropomorphizing everything (I do not think that my childhood Teddy Ruxpin was "alive", of course!), but it does mean not tying (for instance) the ability to speak or make dinner with the ability to have a complex inner life, since it is obvious that there are people who can neither speak nor cook, but who certainly have a very dynamic inner life.
one thing that I am curious about is whether a mind must have some means of getting environmental feedback (and processing it) in order to experience consciousness.
I imagine it would need some sort of input in order to experience consciousness, in much the same way that we need contrast in order to see. That input might not have to be sensory information from the environment, though. Different kinds of abstract data might be enough of a contrast. Those rat neurons might decide that a fighter jet simulation is more fun than a garbage truck simulation.
Post a Comment