The Worth of Wild Ideas
nautil.us
nautil.us
That has served me well many times. It's not a panacea, but rather it's just a form of brainstorming. And it's fun!
Interesting... from my perspective, it seems more pseudoscientific to claim that there is something unique about carbon-based lifeforms, which allows them to be conscious and prevents other things from being conscious.
Like, is the universe supposed to check for the presence of carbon before it decides to endow a computation with a subjective experience? That seems wack.
In any case, I don't understand why there are so many people interested in consciousness, and so few people interested in sensation. Suppose I've been drinking all night, and I'm out cold on the floor, unconscious due to intoxication. You walk up and give me a kick. I'm still unconscious, but I might respond by yelping or rolling over. I won't remember anything about the incident in the morning, but it still caused me pain.
Despite my being unconscious, there is still some process which generated pain when you kicked me. Why is it that whatever occurred in my nervous system generated pain rather than pleasure? For example, there are some neurotransmitters and receptors which are associated with pain, and others which are associated with pleasure -- is there something about dopamine which makes it intrinsically pleasurable? Or is it easy to imagine an organism that evolved to use dopamine to signal pain rather than pleasure? And if the latter -- which strikes me as more likely -- what lever is the dopamine actually pulling? What is actually going on in the brain's "pain center" or "pleasure center" which causes pain or pleasure?
It seems to me that solving the puzzle of sensation should be considered a prerequisite for the problem of consciousness, because consciousness is mix of information processing + varied sensations.
pain is a signal of a potential thread to well-being.
How can it be that some chemistry in my brain's pleasure centers leads to a sensation of pleasure?
(This is an analysis, not experimental truth.)
Pain nerves do at least three things:
1. Generate a normal sensory signal. "Your hand is hot."
2. Trigger a common pain sensory signal. "Your hand is too hot."
3. Trigger the adaptation of a pain prediction model, with a pain prediction signal. The prediction signal is a reinforcement signal that physically ensures: "You do not like your hand to be too hot."
That reinforcement happens in two ways:
4. Trigger negative reinforcement for any action predictive of pain (initiating or maintaining). "You are very motivated not to touch or hold that hot plate!"
5. And positive reinforcement for any action that is predictive of escaping pain (avoiding or reducing). "You are very motivated to wear kitchen mitts before picking up that plate! Or pull you hand away if you accidentally do!"
Which results in motivation:
6. We automatically choose actions that avoid pain. We automatically choose actions that extinguish pain. It would take counter-motivation action NOT do these things.
Pain prediction, and therefore reinforcement, are transitive:
7. Any state that predicts a state that predicts pain, becomes recursively predictive of pain. "You don't like your hand to be too hot -> holding things that are too hot -> holding hot plates."
8. Any action that is predictive of a state change that is predictive of a pain change, gets recursively reinforced (positively or negatively). "You are motivated to avoid touching hot things -> from touching hot plates -> to wear kitchen mitts."
And the dual of all those effects for pleasure.
The transitive/recursive learning helps us to avoid pain, and seek pleasure, at a very high indirect level. For instance, the pleasure of drinking wine or eating sushi, isn't just the physical pleasure of drinking and eating when we are thirsty or hungry, but our bodies prediction that those things will make us feel good over the next hour or so after consuming them.
Now add self-awareness.
When you consider things predictive of pleasure, you feel the the prediction signal, reinforcement, and motivated for that thing.
This is "wanting" and "liking".
When you are in a state producing the pleasure signal, you feel that signal, continued reinforcement and motivation to stay in that state. This is increased if the state is also associated with future predicted pleasure (i.e. sex feels good physically, but is also predictive of relationship stability, validation, etc.)
That makes the experience of "pleasure" far more complicated and meaningful than plain sensory signals.
When our body or mind turns off the pleasure, especially in a way we predicted it would, our motivation stops and we feel ready to move on.
That is "satisfaction".
I think "happiness" could largely be a state where we have high confidence of predictive high level pleasures going forward, or at least in the moment. I.e. peace, success, positive social situation, lack of worries, interesting challenges, etc.
I agree that it's useful to feel pain and pleasure because this adjusts our future actions. But how could sensation actually be implemented? How could I look at equations for one or more chemical reactions and determine whether they happen to generate pain or pleasure?
I don't think the usefulness explains the implementation, because we still feel pain and pleasure in situations where it's useless or harmful to feel them. E.g. it's harmful to feel pleasure from addictive drugs, but people do so anyways.
The experience would correlate to whether a self-aware system had a subsystem similar to our pain and pleasure subsystem.
So you would look at an entity's information/learning system design.
--
As for the experience of pleasure or any other feeling, we respond to our accessible signals without any awareness of HOW we are aware of them or how to describe them - except for associations with other stimuli. Our signals run on "hardware" (neural or electronic) below our ability to monitor.
CPUs have access to registers and RAM, but not their circuits. We have access to neural signals, but not the neurons, transmitters, or hormones activating them.
Our "feelings" are complex mixes of associations, which we recognize, but can't separate into constituent signals and associations. We don't have hardware for that.
That inability to identify our feelings components, gives them their mystery.
--
Here is an interesting experiment: find some nondestructive way to cause yourself some pain, like pinching.
Pinch yourself hard somewhere, but override any avoidance response by focusing on the stimulus instead. Purposely "listen" to it, mindfully learn from it. Frame it as just another sensory signal your are curious about.
You can largely inhibit pain avoidance and reinforcement.
It's a good trick for to reduce chronic pain, or exercise through pain. Every time you start feeling pain avoidance, refocus on it again to tame it.
The brain interprets those signals.
Dopamine is not a “pleasure” chemical. It’s more like neural lube that allows your brain to train on specific patterns And motivation setting and frankly a whole bunch of things.
Actually, this is making me rethink my earlier statement about carbon-based lifeforms -- perhaps whatever chemistry capable of causing pain or pleasure is a strict subset of organic chemistry.
How do the spots work?
Drugs are probably better suited to mess with these than surgeons.
I think you’re asking though if something is intrinsically pleasurable to an organism. The answer is no. But all life has generally evolved to sensibly reward useful things and discourage bad things. So most brains are designed to interpret specific inputs as pleasurable. It’s not inconceivable that with advanced technology we could design an organism that experienced pleasure from whatever sensory inputs.
The fundamental "problem" of consciousness is that systems are deterministic (and/or random), so where is that special spark? Surely a machine can't be conscious, because they don't have that special magic that we do. But when we look closely at our own bodies and brains, they function a lot like machines...
Consciousness is an overloaded concept that folds a lot of things in. Legal culpability, rights and privileges, communication, connection, meaning.
You can easily partition your life in such a way that (percieved) consciousness emerges in things other than people. Places, animals, concepts, ritual behaviour can all fit the criteria, if you let them.
As computing people, distributed consciousness should not seem any weirder than microservices or docker containers.
All that remains is an unanswered question. That special spark is what we are trying to pin down and explain and understand, it might not be special or a spark, but we don't know how else to describe it right now.
do they? Or are you just seeing the aspects of ourselves that we've been trying to place on machines since forever. Or personification, the human aspects we put on machines to relate to them, a car needs fuel, humans need fuel, we're both machines, etc.
Our biological systems are much closer to random than deterministic.
However, they are not purely random, either. Things are very much biased in normally beneficial ways.
And that's before we start talking about processes that actually depend upon quantum mechanics to operate properly (see: photosynthesis).
It's great to see this from a leading voice in the area (Anil Seth).
It's also nice to see a title that matches the content of the article.
How do they define consciousness?
This doesn't include things like "awareness", "cognition", "sentience", or anything else. It's simply "the experience of yellow" as distinct from the physical reactions caused by incident light of a particular wavelength.
In Nagel's terms, "what is it like to be a bat?", not merely perceptually but sensationally.