On Determinism, Free will, AI, artificial life, and simulation theory
Text generated by OpenAI’s LLMs in this article is written in cursive, with a quote from each of the Discord users that participated in the conversation, everything else is written by myself (nixtoshi#2359)
1. Preamble:
[nixtoshi#2359]: I love the comments on this thread https://news.ycombinator.com/item?id=11824164
And this implementation of Aaronson’s oracle: https://roadtolarissa.com/oracle/
I think I’m starting to finally realize that what we call “freedom, god, chaos” is simply phenomena that hasn’t been explained/modeled yet.
So freedom is probably just that instinct, that leap of faith when in doubt, the “bet on red function”, or the “predict right” if not modeled yet, or the “default to motion”, function.
The word “random” which some correlate to “freedom” seems to also have those origins: Gallop, great speed, impetuous, headlong rush, and made without a method.
Maybe to a sufficiently advanced observer “a god’s view”, humanity or the planet is entirely predictable. Only realized this after consciously getting 0%, 100%, 62%, 52%, and 50% at Aaronson’s Oracle on purpose by understanding the algorithm behind it.
Without understanding it beforehand, my actions were fairly predictable by the computer running this algorithm, sitting at around 60–62% predictability over hundreds of key presses.
When I focused and really tried to be equally predictable and unpredictable by the machine, I achieved 50%, but regressed to 52% over hundreds of presses.
When I researched the algorithm and used a countering method, I realized that any number could be sustained against the oracle, changed on the go, and even a 0% predictability was possible.
In response to this observation:
[hiton#8575]: Would you consider yourself a determinist? And if so, a hard determinist or a soft determinist?
[nixtoshi#2359]: I just read more about it, I don’t consider myself a determinist, philosophically I side more with utilitarianism, if it serves human ends, I believe in it. I also like science because it’s utilitarian in a lot of cases.
Determinism seems too perfect unlike the real world which tends to be more chaotic or at least not perfectly modeled/predictable/describable yet.
I do like hard determinism as a goal even if impossible though, encourages trying to make the world more predictable/explainable in simple terms, or more predictable/computable.
For now, I like the thought of freedom/agency, even if it was fundamentally false at the core engine of reality or an illusion that living things believe in.
[RulerOfMeasurement#9378]: I would consider myself a hard determinist with the caveat of computational irreducibility.
[GPT3.5] Computational irreducibility:
- Computational irreducibility is a concept proposed by Stephen Wolfram, stating that many physical systems are too complex to be effectively measured or predicted using only initial conditions.
- This inability to “shortcut” a system’s behavior in a simple way is termed computational irreducibility and highlights the limitations of existing mainstream science.
- Observation and experiment are necessary in cases of computational irreducibility, as no easy theory can fully explain complex behavior.
- Some less complex systems can exhibit simple and predictable behavior, allowing for approximations, but more complex systems remain computationally irreducible and unpredictable.
- Computational irreducibility has been linked to the concept of Compatibilism and may provide insights into free will.
[dantheman3642#9507]: Exactly so even though the universe might be deterministic I don’t think that means we’ll ever be able to predict it in all of its full determinism.
Counterfactual thinking
[dantheman3642#9507]: I also think that even though it’s a deterministic process, us and a few other higher organisms uniquely have the ability to create counterfactuals. Where we can create representations of alternate futures in our head, and base our decisions on which ones we want. In this sense, I think counterfactuals would be a useful correlate to what a lot of people mean when they talk about free will. And in this sense it would be true to say that humans have more free will than other organisms that we know of./ left here
[RulerOfMeasurement#9378]: When the trajectories of three point masses in a void can’t be accurately computed for any appreciable amount of time it would be hard to see how this wouldn’t be the case.
[GPT4]: Counterfactual thinking:
1. **Definition**: Counterfactual thinking involves considering alternative realities or outcomes that didn’t actually occur. It’s a type of mental simulation where individuals imagine what could have happened under different circumstances.
2. **Function**: Counterfactual thinking serves a number of important functions, including aiding in learning and decision-making. By imagining different outcomes, we can analyze our mistakes, learn from them, and potentially make better choices in the future.
3. **Types**: There are two main types of counterfactual thinking: upward and downward. Upward counterfactuals involve thinking about better outcomes that could have happened (“if only I had studied more, I would have passed the exam”), whereas downward counterfactuals involve considering worse outcomes (“at least I didn’t fail all my subjects”).
4. **Impact on Emotions**: Counterfactual thinking can significantly impact emotions, often leading to feelings of regret, guilt, or relief depending on the nature of the imagined alternative scenario. Upward counterfactual thinking can lead to regret or disappointment, while downward counterfactual thinking can bring relief or gratitude.
5. **Role in Behavioral Change**: Counterfactual thinking can be a catalyst for behavioral change. The regret experienced through upward counterfactual thinking can motivate individuals to make positive changes, while the relief experienced through downward counterfactual thinking can reinforce current behaviors as individuals realize how much worse things could have been.
[nixtoshi#2359]: We will tend to enter into one of these 2 states:
Desire for a better state (stress), or relief from not being in a worse state which in humans might be correlated with ownership/economics/social circle or other hierarchical positioning. I remember reading that in chimpanzees, individuals at the bottom of their social structure, and individuals at the higher end of their structure, both experience the highest amount of stress in the form of more cortisol in their bloods.
[RulerOfMeasurement#9378]: My thermostat sees the world as it is, finds it lacking, imagines a different way it could be, and takes steps to change it. Free will thermostat is best thermostat.
2. Question:
Asked by [hiton#8575]: For those of you who are hard determinists and believe free will is not compatible with hard determinism (or who are familiar with the theory):
Are there any practical consequences to this in life? Wouldn’t this mean that even our reactions to conversations about determinism were, in fact, pre-determined (or rather, that no alternatives exist)?
Would this mean that we also have no choice but to continue experiencing the world as if we have free will?
Conversations about determinism are interesting to me and I tend towards being a determinist, but I find it hard to see any consequences from this.
In the strictest possible sense of determinism, the ‘choice’ I made to, for example, think about and write this post out, is just a predictable hallucination that had the natural consequence of resulting in text…
I wouldn’t be able to actually change anything, including acting as if people had free will, and neither would anyone else.
So basically, if hard non-compatibilist determinism were true, nothing would change, because nothing would be able to change.
Am I understanding the theory correctly?
3. Answer:
[nixtoshi#8575]: This paper touches on these questions. The conclusion goes through it quickly: https://www.nature.com/articles/s41599-020-0494-4
The answer from a hard determinist would basically be that yes, “nothing can be changed” as in everything is following its pre-determined course and can’t be altered, or put differently “everything is changing non-stop” according to pre-determined rules.
Thinking about this itself, wouldn’t change anything about the basic physical laws moving the universe either.
The fact that some people figure out these physical puzzles and use them to alter reality or better position themselves in it, would have been pre-determined as well, and not chosen. So the tenant that “free will” exists as in “electing the reality you want to be in with 100% accuracy” would remain false to a determinist.
4. Unpopular belief
Human law assumes that free will exists, and hence looks for responsible and accountable human parties to blame for natural phenomena or phenomena with human intervention (such as a house that was built on supposedly sound principles, and then fell).
Free will also forms part of a lot of people’s personalities and political orientations, so I can see how determinism, which includes humans as part of natural phenomena and basically turns them into natural automata, is unpopular.
5. How to prove determinism
When I think about determinism I think about how to prove this concept/theory fully at 100% accuracy, which may not be possible at 100% accuracy, so we run approximate simulations that are usable enough to achieve some end. I also believe that there is a tendency in chaotic systems, where the bigger the body of study, the more predictable it becomes (for example it’s easier to predict the earth around the sun with a high degree of accuracy, than it is to predict a quantum state, or a person’s behavior with the same degree of accuracy).
What do you think GPT-4?
[GPT-4]: Determinism is a philosophical concept related to cause and effect. It is the belief that every event is necessitated by antecedent states of affairs, and that given those states, nothing else could possibly happen. In simpler terms, determinism means that every event or action is the result of previous events or actions, and could not have happened any other way.
However, determinism is very difficult to prove conclusively due to its very nature, as it is dependent on the entirety of prior events. That being said, there are several areas of evidence often discussed in support of determinism:
1. **Physics and the laws of nature**: Laws of physics such as gravity, thermodynamics, electromagnetism, and quantum mechanics operate with consistency and predictability, which suggests a deterministic universe. However, quantum mechanics also introduces elements of uncertainty and randomness, leading to interpretations like the Copenhagen interpretation or many-worlds interpretation, which challenge strict determinism.
2. **Causality**: The principle of causality, stating that every effect has a cause, is a key support for determinism. However, proving that every single event has a cause is again, practically impossible.
3. **Neuroscience and psychology**: There’s emerging evidence from these fields suggesting that our actions might be determined by the state of our brain before we’re consciously aware of our decisions. However, this doesn’t prove determinism across all aspects of existence, only that our perception of free will may be more limited than we think.
In conclusion, proving determinism is a complex task. While there’s evidence to suggest a deterministic universe, there’s also evidence suggesting randomness and unpredictability. Determinism is a philosophical position rather than a scientific fact that can be empirically proven. The debate between determinism and free will, or between determinism and randomness, is ongoing and one of the most fascinating in philosophy and science.
6. My personal opinion:
I don’t agree with the conclusion of the paper at the start of the conversation, because I don’t believe in determinism.
I prefer to think that there is some free will, or that free will is our actions over uncertain things, maybe even being a living thing is inherently having this pseudo-free will inside a universe that is predictable to a certain extent, and unpredictable in some other scales of matter, or timeframes.
To a living thing, after all the methodology, models, and reasoning have been considered, the future is usually not perfectly predictable, so we take leaps of faith on imperfect information, I call these leaps of faith, the “freedom” of living things, or our ability to take decisions and biases based on imperfect information.
From the viewpoint of a person that thinks deterministically, I see the logical flaw in thinking like this, but I simply prefer it, I find it more appealing to think that there is some free will rather than not, and encourage trying to turn the world into a more predictable world by making it more deterministic through models and environmental modification, rather than simply assuming it can’t be done, or believing that there is a level of chaos that can’t be predicted/described, because being proved wrong on this, is highly utilitarian to humans.
7. Artificial life
I also think that intelligent artificial living things, and even better, an AI that is indistinguishable from human intelligence might count as a proof that humans can be virtualized in simulated worlds and that we may simply be more complex and larger scale versions of these simulated, computer-run, computer simulated intelligences.
A replica of human intelligence or a smaller living thing, might also be possible, even if we don’t yet have the technology to make them at scale.
For instance, it’s probably possible to build small insect robot in a controlled environment that consumes energy, resources from its environment, reproduces and self-repairs.
I also see how it would be possible to endow these beings with a hard to predict “free will” function, machine learning algorithms, and the capacity of self-augmentation.
To a lot of people that would probably count as human-made artificial life.
A highly unpredictable randomness generator could for example be based on a combination of ambient noise, the level of light in the environment, or other unmodelled phenomena from the perspective of the automaton, like small variations in the speed of some computations, etc. A lot of these sources of entropy are unpredictable enough, that even we use them to secure our secrets, information and money.
A more efficient “free will” algorithm that creates faster learning to achieve some biological objectives, would be what I call the “bet on red function” which is simply taking the same choice every time there is an unknown, unmodeled scenario, in order to save time/energy, to increase learning speed.
However, this behavior would be entirely predictable to an attacker with perfect historical knowledge of the automata’s actions, as proven by my meddling with Aaronson’s Oracle.
A more complex free will algorithm would consume more energy and be less efficient, but would result in a lot more unpredictable behavior to an attacker with perfect historical knowledge of the automata’s actions up until that point, because, even the attacker can’t perfectly simulate/control the source of randomness that the automata is using, unless the attacker captures the automata and captures it in a perfectly controlled environment.
This perfect environment however, could become more chaotic and less predictable when the automata’s attacker observes the automata’s state, since it might risk introducing unaccounted variables.
In this sense a perfectly predictable “artificially made” automata requires an entity with perfect knowledge and control of its environment.
Most human creation is more chaotic by default, because we don't have a perfect understanding and control of our environment.
8. Reality as a computer simulation
There is also the simulation hypothesis.
While it’s possible that we live in a simulation, it might not be that relevant because:
It’s possible that we can’t prove it. This would assume that the security systems designed to hold us in place are beyond anything that we can break (think of a huge glass container holding bacteria).
And if we did prove it somehow, what is our course of action then?:
- Crack or escape the simulation into the originating reality:
1.1 We can escape, but maybe the people/entities running the simulation don't want us to, which might cause our elimination/destruction (dead end).
1.2 We do escape the simulation, like a hamster that goes out of his terrarium or a fish that escapes its aquarium.
1.2.1 We are returned into our simulation (return).
1.2.2 We die in base reality due to a lack of energy source (dead end).
1.2.3 We simply observe the outer reality instead of being able to change it (Discovery: a peak into another realm of existence).
1.2.4 We become a non-coherent data/crash/glitch which is discarded, therefore we disappear, and the simulation is returned to a previous state (dead end with time travel/correction). - We enter a different stable realm, or another simulation (transplant). And are able to continue our existence outside the simulation we spawned in, maybe we continue our existence normally, maybe not (we make it alive: Therefore we have some continued existence).
- Goal achieved, potential reality halt: By exiting the simulation, we successfully complete some sort of function/objective (which leads to a potential dead end, or continuation). By completing a function, our simulation might seize to exist as the simulation might be stopped, or we might win a prize, or we might simply wake up in a different realm of existence, but this isn’t guaranteed).
In this 3rd scenario the possibilities become too broad, almost as broad as our imagination of what can be programmed in our own simulations. So religious theories, alternative theories of everything and more, could have a chance of being true by chance/coincidence, or even through actual information transfer from the origin reality into the minds of certain people who wrote these alternative theories of everything. - Base reality limit. It’s possible we might simply exit the simulation, but enter another simulation. Like a maze inside a maze, or a game inside of a game, inside a game. Until we enter base reality which can’t be escaped from, and from which all of the other realities originate.
- Reality as an infinite loop. It’s also possible that reality, regardless of wether it’s simulated or not, is an infinite loop that can’t be escaped from (infinite continuation outcome).
- Infinitely looping simulation. It’s also possible that one simulated reality leads to others, but even when all simulated realities are mapped, you always end up exactly where you started meaning that there would be no more simulated realities to explore, and reality would be an infinitely looping simulation, parametrized by different rules, states, or sizes, but ultimately infinite, never ending, and possibly unchangeable (infinite unbreakable continuation outcome).
9. Conclusion:
For matters of practicality, simulation theory, and determinism are strictly theoretical for now, and might not have any effect in our daily lives until we prove them or find utilitarian use for those theories, through new discoveries of the nature of reality, such as quantum computing.
