You pose the options as (1) your choice was predetermined by the purely physical (atoms, laws) (2) your choice was freely made by your agentic self. Have you considered that your choices may be (pre)determined by psychological causes? That there may be reasons, conscious or unconscious, as to why a particular impulse arises in your mind and is allowed to issue in an action? Certainly I don't feel as if my own decisions arrive unaffected by thought, feeling, unconscious association, etc.
I also wonder if your sense that "you could have decided differently" really derives from a feeling that the dominant causes leading to your choices are within your mind (and therefore relatively unaffected by external causes), rather than a feeling that your choices have no cause but an impulse of the will that is itself uncaused.
Yeah, that's an interesting point. I would say that many choices I make are affected by psychological factors, which are probably reducible to atoms/laws (like chemicals in my brain). What I'm curious about is if my choices are entirely reducible to these psychological and other physical factors, chemicals in my brain, movement of atoms according the laws of physics, etc, or if there is any part of the choice being made by my conscious mind in a way that can't be accounted for with the standard physical model of the world.
have a strong intuitive feeling that my mind acted as an agent, made a choice to lift up my right hand, and really could have chosen not to if the universe were re-run
Speaking of fuzz and muddle, that's kind of running together two things: libertarian free will with some notion of agency.
Different definitions of free will require freedom from different things. Much of the debate centres on Libertarian free will, which requires freedom from complete causal determinism (and therefore, freedom from inevitability). Libertarian free will requires leeway, or multiple possibilities at a moment in time. That's often expressed as being able to act differently if the history if the universe were rewound. Compatibilist definitions of free will only require freedom from compulsion, and allow free will to exist in a deterministic universe . A third set of concerns is free will is a form of conscious control,top down control or initiation of action by an inner self
The idea of conscious control doesn't have to be entirely abandoned by naturalists. You don't have to appeal to ghostly inner selves, because you can appeal to higher brain functions.
By “free will,” I mean the question of whether conscious minds have metaphysical agency, in the sense that they can cause physical events and make decisions, in a way that is not accounted for by the standard physicalist model of initial conditions, deterministic laws of physics, and randomness.
Why does it have to be metaphysical? Is the problem for free will the physical or the deterministic? Physics isn't necessarily deterministic.
According to science, the human brain/body is a complex mechanism made up of organs and tissues which are themselves made of cells which are themselves made of proteins, and so on.
Science does not tell you that you are a ghost in a deterministic machine, trapped inside it and unable to control its operation. Or that you are an immaterial soul trapped inside an indeterministic machine. Science tells you that you are, for better or worse, the machine itself.
Although I have used the term "machine", I do not intend to imply that a, machine is necessarily deterministic. It is not known whether physics is deterministic, so "you are a deterministic machine" does not follow from "you are entirely physical". The correct conclusion is "you are no more undetermined than physics allows you to be".
So the scientific question of free will becomes the question of how the machine behaves, whether it has the combination of unpredictability, self direction, self modification and so on, that might characterise free will... depending on how you define free will.
There is a whole science of self-controlling machines: cybernetics. Airplane autopilots and , more recently, self driving cars are examples. Self control, without indeterminism is not sufficient for libertarian free will, but indeterminism without self control is not either
All of those things can be ascertained by looking at a person (or an animal or a machine) from the outside. They don't require a subjective inner self... unless you define free will that way. If you define free will as dependent on a ghostly inner self, then you are not going to have a scientific model of free will.
If an AI responds to something you say, it is responding the only way it possibly could, given the inputs it receives, the model weights, and any random numbers that are used in the computation. If the universe were re-run with the AI receiving the same inputs, with the same weights and random numbers used, it could not respond any differently. There is no space for the AI to *choose *to respond differently in a way that isn’t accounted for by the physicalist model of reality. We can confidently say that AIs don’t have this type of free will.
But it would be easy to convert it so that it used a source of quantum randomness instead of PRNGs. If that is the only requirement for a freely willed AI, then you can build a freelly willed agent out of atoms, and no spooky metaphysics is required.
Or maybe human minds don’t have this type of free will, and it’s just an illusion on top of a complex processing of physical matter that is ultimately either deterministic or random
If the physical matter is using a mixture of determinism and randomness, why would it have real free will?
Or maybe human minds don’t have this type of free will, and it’s just an illusion on top of a complex processing of physical matter that is ultimately either deterministic or random
What does a quale of irreducibility feel like @Mitchell_Porter
"Unaffected by" and "influenced by" are not the only options either. all senaible libertarians admit there are influences. ?
How likely is it that free will emerges from the very randomness which you cite and discard as "The output of an AI is completely determined by its inputs, model weights, and any random numbers that are used in the computation"? IIRC the LLM chooses the next token randomly after it calculates the energies and probabilities as
I suspect that the closest equivalent of such a situation which the humans can encounter is facing a choice between two alternatives of similar expected distributions of utility-like functions, or between a good and bad alternative, while the distributions themselves are governed by human knowledge, intuitions or even emotions experienced at the time via processes whose description cannot be contained in this margin depends on mechinterp applied to the human brain (or, if the dilemma is posed before an LLM, to the LLM).
I think LLMs typically use pseudo-random numbers for these calculations, and so are still ultimately deterministic given the seed of the pseudo-random number generator, such that if you re-ran the AI on the same inputs, with the same weights and PRNG seed it would give the same output, with no room for a metaphysical agent to be exerting free will.
We might be talking past each other, possibly my fault if I'm using a non-standard definition of "free will." But what I'm trying to get at isn't about decision theory and the way individuals choose between different options to maximize a utility function, it's instead about the relationship between physical matter and consciousness, and the strong intuitive feeling that my subjective consciousness is an agent that is causing physical events in a way that is not reducible to deterministic physical causation or randomness. And wondering if this is real or an illusion, and noting that AIs certainly do not have some unexplained metaphysical property like this, because their processing of inputs into outputs can be fully traced from end to end. Possible also that your point was about this and I just didn't correctly understand it. Anyway thanks for the comment!
I think discussions of free will often become fuzzied and muddled in a way that makes them less interesting. For example, sometimes there's a debate around free will that goes like this:
Ok, I guess maybe that's still interesting in a way. Actually it reminds me a bit of The Godfather, where Michael Corleone starts off as an upstanding citizen who wants nothing to do with the family business and [SPOILER] ends up as a mafia boss who orders the murder of his own brother. At no point does he suddenly decide to become an evil mafia boss. His transformation follows a chain of events, where at every specific point he seems to be acting reasonably and doing the only thing that he can do, without any real choice in the matter. Did he have free will in his choice to become a mafia boss?
So maybe discussing free will in this sense makes for a good legal theory debate or movie plot. But it's not what I'm talking about when I say "free will."
Here's what I'm talking about -- I lifted up my right hand just now. That is a physical event that just happened. Could I have not just lifted up my right hand? If the universe were re-run, starting from 30 seconds before I lifted my hand, with all of the atoms in the same place, with the same properties and laws of physics, and all random events leading up to the hand-lift happening exactly as they happened, could I have not lifted up my right hand? Or was I bound to lift up my right hand by the state of all the atoms in the universe, the laws of physics, and the outcomes of any random events that happened?
I have a strong intuitive feeling that my mind acted as an agent, made a choice to lift up my right hand, and really could have chosen not to if the universe were re-run. But I'm not sure about this, and this is the "free will" question that I'm curious about. By "free will," I mean the question of whether conscious minds have metaphysical agency, in the sense that they can cause physical events and make decisions, in a way that is not accounted for by the standard physicalist model of initial conditions, deterministic laws of physics, and randomness.
It sure feels like this is the case. But again, I'm not sure. This type of free will, where conscious minds are causal metaphysical agents, is inconsistent with the standard physicalist model of the universe. I'm not the first person to notice this, and philosophers like Jaegwon Kim have written extensively about it. So maybe this is a point against the existence of free will. But I have such a strong intuitive sense that I possess this type of free will, that its incompatibility with physicalism seems like it could also be a point against physicalism as a model of reality, at least as we currently understand it.
I'm not sure. One thing I think I can say confidently though, is that AI does not have this type of free will. The output of an AI is completely determined by its inputs, model weights, and any random numbers that are used in the computation. If an AI responds to something you say, it is responding the only way it possibly could, given the inputs it receives, the model weights, and any random numbers that are used in the computation. If the universe were re-run with the AI receiving the same inputs, with the same weights and random numbers used, it could not respond any differently. There is no space for the AI to choose to respond differently in a way that isn't accounted for by the physicalist model of reality. We can confidently say that AIs don't have this type of free will.
If you're unconvinced by this, it could be a fun exercise to prove it to yourself by building a simple AI model from scratch, so that you can see how it works internally and confirm that it is a non-magical, non-metaphysical, surprisingly simple math calculation that takes inputs and gives outputs. Andrej Karpathy has a great tutorial series where he teaches how to build a simple AI model completely from scratch in Python. Or if you'd prefer to learn from someone less-cool, not-as-smart, and worse-at-teaching, I also have a tutorial series where I show how to make a simple AI model from scratch in the Julia programming language. Anyway, when you build a simple AI model from scratch and see what's happening at every step to get from inputs to outputs, you can conclusively confirm that its intelligence is not something mysterious, and that it is based entirely on deterministic calculations, possibly involving added randomness, but definitely without the kind of free will that would involve metaphysical agency.
I don't know why I'm writing this down, or if anyone will find it interesting to think about. Maybe it's because I'm interested in the question of what consciousness is, and if AIs are conscious or could ever be conscious. The question of AI consciousness is interesting both as a fundamental question about the relationship between matter and consciousness, and also as an important ethical question about whether we should care about the wellbeing of AIs. But I have no idea how to get at the question of AI consciousness, or the hard problem of consciousness in humans, so maybe I'm just trying to get at something more tractable. If conscious human minds have this property of free will / metaphysical agency (which again, I'm not sure about), and we can conclusively say that AIs do not have this property, then maybe that's a point against AIs being conscious even if they can convincingly imitate human minds.
Or maybe human minds don't have this type of free will, and it's just an illusion on top of a complex processing of physical matter that is ultimately either deterministic or random. I don't know. Anyway, today I was just thinking about whether human minds have free will in this sense, and what having it or not having it would tell us about the nature of minds, and if there could be conscious AI minds.
Interested to hear what people think about this. Also, are there any physicalist models of reality that support conscious minds having this type of free will? I think I've heard John Wheeler's observer-participatory universe idea and Chris Langan's CTMU associated with this, but I haven't been able to grasp them enough to evaluate them (if anyone has looked into them, interested to hear about it). So curious to hear if there are physicalist models of reality that are compatible with this intuitive sense of free will in the true sense of being metaphysically causal, and interested to hear what people think about this question in general.