An example of misalignment with information technology is social networks. At first they were hailed as a major benefit to humanity, bringing us closer together online. But then we found out the downsides after social networks had become very popular and widespread. Turns out the algorithms maximize engagement to sell ads, which often ends up being whatever makes people angry, among other things.

What makes alignment a hard problem for AI models? Because they are based on gradient descent using giant matrices with varying weights. The emergent behaviors are not well understood. There isn't a science to explain the weights in a way that we can just modify them manually to achieve desired results. They have to be trained.

So it's a difficult task for OpenAI to add policies to ChatGPT which can always prevent it from saying anything that might be considered harmful. There are are always ways to hack the chat by prompting it to get around those policies.

This also raises the discussion of who gets to decide what is harmful and when the general public should be protected from a language model generating said content. If future advances do lead to a kind of Super AI, which organiation(s) will be the ones aligning it?

That would be an anti-realist interpretation. Sean Carol is a realist about the wave-function, so he thinks there literally is a multiverse, at least from after inflation until heat death of the universe.

Reasonable as in making sense within the MWI interpretation. MWI needs to be self-consistent and not have to introduce anything from outside the wave-function to make things works. So as long as observers and observation can be understand as parts of the universal wave-function, it's reasonable. I still have questions, though.

But when inflation ends, the universe reheats into a hot plasma of matter and radiation. That actually does lead to decoherence and branching — Squelching Boltzmann Brains (And Maybe Eternal Inflation) - Sean Carroll

That's informative and interesting. So once inflation ends, the multiverse begins, until De Sitter space, when there's nothing left to decohere and make observations. Then all is just superposition.

That sounds mostly reasonable, but the branching part based on something making observations still bothers me a bit. What is the branching mechanism? Perhaps I should have started with that question instead.

So, coming back to this thread after many days away, Sean Carol has stated a solution to the Boltzmann Brain problem is that there won't be any observers in De Sitter space to cause decoherence under the MWI. Boltzmann Brains are thought to be the results of quantum fluctuations over an infinite amount of time after the heat death of he universe, but the wave function is deterministic, so as long as there are no decoherent branches, there's no sense of fluctuation.

It's still weird to me that the observer is a necessary component of making sense of MWI, since decohered branches are still in universal superposition, which is what infinite De sitter space will become, except without the decohered observers.

Keith Frankish's illusionism argument. That the brain is performing the equivalent of a magic show, tricking us into thinking there's something about consciousness that turns it into the hard problem. I can't be sure exactly what his argument amounts to. He seems to be denying the phenomenal aspects of consciousness, since those are what leads to the hard problem. So I guess he's arguing for a functional account with the added twist that are brains trick us into say things like the "redness of red", or there's something it's like to be a bat, which we can't discover with neuroscience. It only seems like we have qualia.

Chalmers has said that if there is a dissolution of the hard problem, the meta-problem of explaining why we think there's a hard problem has to first be addressed. Frankish attempts to do that. I just don't know whether it seems like I'm phenomenally conscious is different than actually being conscious in the hard sense.

He has a very interesting idea on how to put MWI and wave-function collapse interpretations to the test. Assuming we can build a conscious AGI quantum computer.

The rest of what he says sounds similar to Sean Carrol's arguments for thinking MWI is likely correct. That it explains the interference patterns seen in experiments when a measurement isn't made, that there's no clear dividing line between the classical and the quantum, and entanglement means all the particles making up classical stuff should be quantum. And that any other interpretation would have be at least as complex as MWI, and probably more so.

That being said, my understanding is that the probabilities we use to calculate the likelihood of what to expect when a measurement is made still needs to be derived within the Schrodinger equation in a self-consistent manner without adding it in post hoc, since the wave function is supposed to describe the universe we live in, if MWI is true. So deriving the Born rule within MWI is an ongoing project.

As an aside, I was listening to Sean Carol being interviewed, and he made it sound like the very low probability events didn't happen, at least not over the time period since the Big Bang (not nearly enough time had passed). However, they should be happening in the sense of being expressed as superpositions by the universal wave function in MWI.

So there's some states where the particles of the rock are located in other parts of the universe, under the understanding that a particle's position ranges over the entire universe, with most of the positions within the place we'd expect to measure them. But there still would be a few spread out everywhere else.

There should even be some human-like observers seeing a rock teleport some distance, or just vanish into being spread out all over the place, and all sorts of scenarios in between, even if it's a vanishingly small subset of observers.

I never said I wasn't crazy. Or didn't make typos, whoops.

I don't know if that would be a logical error. I'm guessing the strong bias towards believing that we're all the same has to do with communication. — frank

It's obviously not the case if you've aware of savants or various neurological abnormalities, which you would hope educated people like philosophers and scientists would be aware of when making claims about the mind.

Yes, my experience is the same as yours. I read other posts from people with aphantasia and they make the same mistake. They think we are walking around with HD movies in our heads. some people do, but I guess they are at least as rare as people with aphantasia. — hypericin

What about when you dream? I would put it more in terms of a VR headset kind of experience, particularly for lucid dreaming.

Some people are really good visualizers. Others can compose music in my head. I have a regular stream of inner dialog. I wonder what you make of Temple Grandin's claims that for autists like herself, their imagination is like the Star Trek Holodeck.

Of course in all this I'm reminded of the certain scientific and philosophical skeptics who mistake their lack of visualization or lucid dreaming for those abilities not existing in other people. That's a kind of logical error whose name escapes me.

Our classical appearance needs to be part of a valid solution to the universal wave function, and nothing says it is not. — noAxioms

Sabine Hossenfelder says it's not:

I would then expect us to find out that bats aren't that different from humans, that all animal mental worlds are variations on a common theme, just like all animals use the same genetic code, and tend to share vast amounts of DNA. Your hemoglobin is quite similar to bats'. We're all cousins. — Olivier5

But that doesn't mean bats or other animals have the exact same set of sensations. We know that can't be true because many birds can see more than three primary colors, and presumably bats have a sonar sensation. Maybe it's a kind of color or sound, but it could be something altogether different as well. And what would it be like as an octopus, where the nervous system is as much distributed in the tentacles, which act semi-independently, as it is in the head?

Even if so, that doesn't mean consciousness is understood functionally, as in we can provide a function which makes a system conscious. If we could, then we would know how to do the same with computer programs and robots. Chalmers criticism is that no amount of structure and function results in an explanation of consciousness. Which is similar to Locke's primary and secondary qualities. Number, shape, extension, composition don't give you the sensations of color, taste, etc. Nagel used that to show the fundamental objective/subjective split in our descriptions. We can't say what bat sonar sensation is no matter how good our science is.

It should be noted though that Chalmers has proposed property dualism on information systems, so he's fine with functionalism as long as there's something additional that connects it to consciousness.

Everett doesn't control how the interpretation develops after him. Sean Carrol, a current proponent of MWI, talks of universes splitting. There's a Universe Splitter app: https://cheapuniverses.com/universesplitter/

Yes, a classical rock takes measurements. If that makes it an observer, then fine. It doesn't need to know about Schrodinger's equation in order to measure a classical world. If you don't count that as an observation, then I completely disagree with your statement above. — noAxioms

There aren't classical rocks or observations in MWI. And yet we make a classical observation, which some call the wave function collapse in other interpretations, every time a measurement is made. You can say the detector or a rock also makes the same observation. But the universal wave function doesn't make such a distinction. Every quantum state is still in superposition.

The point about human observers is we're the ones interpreting the mathematical formalism as meaning reality is this or that. Some physicists, mathematicians and philosophers say the wave function describes the universe. If it does, then the classical appearance of our world needs to be derivable from that equation.

Agent Smith never did really understand what the Oracle was up to. Neither did the Architect until the end. There's several good YT videos that do a deep dive on the trilogy. Some even argue Agent Smith is actually The One.

Anyway, the Oracle, as an intuitive program, recognized that the fight between the machines and humans was just going to continue in the same cycle, so she wanted to find a way forward where they could both coexist in a less combative state. A way for humans and machines to evolve their relationship. To do this, she had to risk everything to force both sides into making peace. The Architect and the machines lose control over Agent Smith, forcing Neo to make a deal they will accept if he gives the machine he's plugged into the ability to identify the Smith virus and eliminate it. The Oracle shows Neo the way by letting Smith turn her into another Smith. Neo must concede the fight so Smith will take him over, allowing the antivirus to take out Smith, and the Architect will then honor the peace agreement, as you see when he meets the Oracle in the park.

As for Agent Smith possibly being the the actual One, you could argue the Oracle lied to everyone including Neo so that she could use Smith to force the peaceful resolution. Neo was necessary because of his special status (somehow both connected to the machine world and humanity), and that needed to be transferred to Smith so he could become a virus.

I updated my comment and added a comment on Kant and the hard problem (that he would likely find it pointless).

This is also how Kant used the term. The noumenon for Kant is an object of intellectual intuition (non-sensible representation of reality).

The difference is that Kant argued that such intuition is a faculty we do not have. — Jamal

Is Kant saying we reason that the real world responsible for our senses is beyond our perceptions and reason? There is a real world responsible for us reasoning and perceiving, but it's unknowable and we can't say anything meaningful about it, only the one of appearances our minds shape from our sensory manifold?

I wonder what Kant would make of the modern consciousness debate. I suspect he would think it's beside the point with both sides making a fundamental error of mistaking the phenomenal physical for the noumenal. There's no point in arguing whether there's a hard problem if it's all phenomenal anyway.

Yes. — noAxioms

Problem is you have to square this with actual observations, which have classical results when a measurement is performed.

Observers as such play no role. Think systems in a state, such as a classic rock at time T. Anything that rock has measured (a subset of what's in its past light cone) is part of the entangled state of that system. — noAxioms

I'll refer you back to what Bohr had to say regarding experiments. Experiments have to be described in terms of the language of performing the experiment, not the mathematical formalism used to model what happens during the experiment. Rocks didn't come up with the Schrodinger equation or the Born rule. Physicists did after observing or learning about experimental results.

No, a world is not a relation with an observer. Not sure where you get this. If you like, you can assign a world in relation to an event-state, but calling the system an observer seems to suggest a very different interpretation. — noAxioms

If there's no observation, there's no world, since as we both agree, a world is a system that appears to be classical. Without an observer, you just have superpositions. Decoherence only matters in this context for explaining why observers don't notice the superpositions.

Yes, and my point is that with physicalism, the question of whether x is conscious will always be open-ended. That suggests the physicalism framework is a dead-end. — RogueAI

We certainly have problems drawing the line on which life forms are conscious. And we can't say what sort of sensations animals with different sensory abilities from us would have. In the far future, there could be Boltzmann brains fluctuating into existence with bizarre mental states that we can't even imagine.

Well, I'm a physicist so I'm going to be biased toward the physicalist/materialist PoVs. I tend to think that property dualism explains things reasonably well, though. — tom111

Chalmers espoused a property dualism in one of his books where any informationally rich system would be conscious. He's more predisposed to finding a universal law connecting consciousness to the physical than just identifying it with certain biological creatures.

Ned Block wrote a paper on the Harder Problem of Consciousness using the android Data from Star Trek to illustrate the problem that we can't tell whether consciousness is tied to our particular biology or functionalism. As such, we have no criteria for deciding whether Data is conscious.

Some people have suggested that the recent machine learning models exhibit conscious behavior. I have serious doubts, and most researchers would probably disagree. But at some point it's likely we will create a machine that's convincing enough where we can't tell. The movies Ex Machina and Her would be good examples of this.

all of whom can be dealt with by any sufficiently armed group of people. — NOS4A2

True, but a sufficiently armed group of people are not likely to be anarchists.

for he’s already been denied for so long the right and means to protect himself that he’s been left a sheep to the wolves, so to speak. — NOS4A2

Sure, but the alternative has been mob justice, which is judged historically to be even worse than the judicial system we've developed over time.

fI’m not positive a group of anarchists are any better at doling out violence and justice than a government, but it’s difficult to see how they can be any worse. — NOS4A2

While true, a lot of that is a matter of scale. Governments can marshal armies because they have a lot of people. They can protect multinational corporations because we have global trade networks. The anarchist faces the problem of what to do when there's lots of people concentrated in areas. It's all good and fine for small groups of hunter/gatherers to be community-based, it's another thing when you have millions of people nearby. There is a tendency for self-organization to occur, and a tendency for some individuals to take advantage of that. Also a need for large-scale organization as services needed to be provided for those millions, and it's a lot more efficient to have highways than a bunch of privately owned roads.

Also, we have a lot of historical evidence for all the wrong-doings of governments, we have less evidence of what our ancestors were up to before recorded civilization. We do know all the other hominids went extinct along with lots of megafauna. Our hunter/gatherer ancestors may have played a role in both. There is evidence of human migrations coinciding with existing populations being replaced.

The reality is there is no ideal solution. Governments are bad because they are run by people, but what alternative is there? Communities are run by people too. It might be a case of what is the least-bad, realistic approach to governance. Same with economics.

bviously I have no problem with people consensually interacting and voluntarily committing to mutual obligations, preferably also without violence playing a role. — Tzeentch

And if people violate those mutual obligations, or wish to be violent? What do you do with Viking marauders or pirates? Warlords, criminal gangs, serial killers, rapists? What about would-be conquerers who are raising an army? It happened in the past. Plenty of rulers conquered their way into power.

Even if we all agree anarchy was morally superior, how do we suppose the world remains in anarchy? It certainly didn't in the past. How would we even ditch thousands of years of government across the planet at this point? 8 billion people are going to live happily in anarchist communities?

Yet if the politicians passed some law or laws you don't agree with, does anything make those laws so special that they have to be obeyed even if they are dumb or harmful? — AntonioP

And what if I want to do things that are harmful to others, because I'm a selfish cunt and don't agree with rules against exploiting others? How does the anarchist deal with that sort of fellow?

The justification for authority starts with all human groups developing rules to follow so they can meaningfully coexist. That means some restriction on freedom. We learn this as small children when older people don't just let us bite, kick, steal and throw tantrums for any reason. Someone has to decide on and enforce those rules. Governments are a way to do this along with administering societal functions like collecting taxes for roads, defense, etc as humans congregated in larger groups.

My response would be, don't try to control people against their will. — Tzeentch

And if their will is to harm others, what then? What if their will is to control others? Or maybe they just want to burn down the nearby forest because they like burning things. Do you just let people do whatever they want? That's not how any society functions.

In fact, almost all quantum experiments are performed without human observation, and it is only well after the fact that the humans become aware of the results in analysis of the data. — noAxioms

Wouldn't that just mean the results could be in a superpositioned state until some human makes an observation? That's the basis of Schrodinger's criticism of the Copenhagen Interpretation, but how would we rule it out?

Experiments show how subjects’ auditory or visual perception is influenced by what they are told. — Joshs

Does that work for colors? Do you think if someone said you would be seeing a gold dress that it would necessarily mean you saw it as gold and not blue? If so, then how did it become a big controversy on the internet with people disagreeing on what color it was? I'm not aware that color illusions work that way. Anyway, it wouldn't matter for insects or birds who can see visual patterns on flowers or other animals invisible to us.

think the wrinkle is in red is a property as we see it. It's as if 'red' is supposed to do double-duty for some ineffable private experience which is somehow known to be the same ineffable private experience for all (an impossible public-yet-private experience). Ryle attacks this kind of confusion in The Concept of Mind, just as Wittgenstein does with his beetles and boxes. — Pie

If this were true, then we'd have no trouble figuring out what sort of colors a tetrachromatic bird sees, or what sort of smells a dog experiences. It also doesn't make sense out of how some people can experience seeing a gold dress and some a blue one.

Take an experiment with that blue/gold dress before anyone knew about it. How would you know that someone was seeing a different color (blue or gold) than you were (gold or blue) until they told you? You couldn't know just by showing them if they're instructed to keep quiet about what the dress looks like.

All of this is rather obvious. We do dream after-all, and nobody can share our dream experience. Many of us have inner dialogs and day dreams. People lie and there's no foolproof way to always tell. Nor can we always know what someone is feeling or thinking.

I don't know how it's possible to escape the conclusion that we do have private experiences. How else would you make sense of the above?

What really....I mean REALLY....is the problem here? — Mww

In typical philosophy forum fashion, nobody can quite agree on the terms under dispute, in part because we have our philosophical commitments to uphold.

hen the ground of the possibility of both, each limited to its own specific domain, but functioning in unison towards a given end, becomes the better option. — Mww

How would that look?

I've been reading Color Realism and Color Science and Color Properties and Color Ascriptions: A Relationalist Manifesto. The first one is a good overview of colour realism and its discontents. — Jamal

I've read Color Realism and Color Science before. It is a good overview. Seemed like a quality attempt at defending color realism, even if I'm prone to disagree.

...and he was cited as an example of these direct realists. IF you have a different source, perhaps you could cite it for me? — Isaac

I didn't cite him. I'm not aware of Locke being a direct realist. Maybe with regards to primary qualities?

You can read a summary about color primitivism and other theories of colors here: https://plato.stanford.edu/entries/color/#RivaTheoColo

Direct realism is a separate but related debate, just depending on what the direct realist has to say about color. As you can see on SEP, there different theories of colors, and a direct realist might choose the one they think offers the best defense for direct perception.

Yes. The idea that seems to be bring presented is that red is some property of an object which produces the response we call 'seeing red'. — Isaac

No, the idea is that red is a property as we see it, not something that causes us to have a response, which could be something unlike color, such as a photon's wavelength.

can find no support for this. Both Locke and the colour primitivist agree that we can be mistaken. So they do not think the world is as it looks to us under proper lighting conditions, at least in the visible light range. Otherwise we couldn't be wrong and both admit that we can be wrong. — Isaac

Under normal conditions, when there's not an optical illusion, and taking into account whatever details about color vision need to be accounted for. The claim is the world is basically colored-in as we perceive it to be.

So it appears even the critics are agreed that the colour primitivists are still assuming colour is a property which we detect and produces the way it appears, not that colour actually is 'the experience of red' in an object. — Isaac

The way it appears would mean the color in our experience. Except that one thing is a property of the object and the other is a perception of that same property.

'm not finding, in the sources you've provided, the idea that any direct realist considers objects to actually have (rather than have a property which causes) the 'experience of red'. Do you have any less ambiguous sources, or perhaps you could explain them more clearly? — Isaac

They certainly don't mean panpsychism. They mean the world is as it looks to us under proper lighting conditions, at least in the visible light range.

If there's no laws governing what can be then all theories are equally valid.[/quote]

No laws or no physical laws? Why do laws have to be physical?

But I think there has to be a reason, otherwise, anything goes - because there are no reasons why this should be a brute fact as opposed to something else. It's a brute fact in virtue of the reason it is the wat it is. — Manuel

Agreed. I also think science without universals doesn't work.

I'm not sure what you're saying here. Are you saying that Humean causation isn't the counterfactual theory of causation? — Michael

It is. I'm disagreeing with it on those grounds.

It's not supposed to. The counterfactual theory of causation just explains what it means for A to cause B. — Michael

As a Humean account. Non-humeans call that contingency.

And perhaps there is a(n unsolvable) problem of induction. How can empirical facts allow for deductive inference? — Michael

Physicists seem to manage. Newton said gravity caused mass to attract. Newton didn't have an explanation for how it worked. So a good example of B always following A. But relativity says it's because mass curves space, changing the trajectory of objects. Now we have an explanation for why B follows A.

You seem to be conflating epistemology with ontology. That we can't know that the universe will always behave a certain way isn't that it won't. — Michael

I realize this. But Humean ontology makes it so we can't know. The problem of induction exists because Hume stated that causality was a habit of thought, not something empirical. That's the skeptical bullet he bit.

I want to climb to the top of that mountain — Xtrix

Since I recently just watched it, reminds me of the documentary Free Solo, about Alex Honnold climbing 3200 feet of El Capitan without a rope in 2017. It's the greatest free solo to date. Alex had the goal back in 2009. He kept a climbing journal, and would spend part of the year living in a van at Yosemite, so he could practice with ropes and other elite climbers. His implementation was going over the route until he was comfortable enough to climb without a rope. He had help planning it out from Tommy Caldwell, the most experienced big wall climber who has made his own incredibly challenging routes on El Capitan with ropes (free climbing). He also has a documentary.

The thing is Tommy said he would never climb El Cap without a rope, because one mistake and you're dead. Part of what makes Alex special is his mental approach. He states that we're all going to die one day, free soloing just makes that very present, and he enjoys doing it because you have to be perfect. Alex had also done free soloing a thousand times on easier climbs. His thing was to get himself into a place where he was no longer afraid, so he could climb a 3200 foot cliff without a rope and not freak out.

Just thought that was a good example of implementing a goal to accomplish something that seemed impossible.

And what does it mean to say that A forces B to happen if not just that if A didn't happen then B wouldn't have happened? — Michael

It means we have an explanation for how A causes B to happen. In physics, the electromagnetic force is the explanation for chemical bonds. Chemistry happens because there is an EM force.

The counterfactual theory of causation is an account of the meaning of causation. — Michael

It is one meaning of causation. It is not the classical meaning. It is a Humean formulation.

But knowing whether or not A will cause B has no bearing on what it means for A to cause B. The counterfactual theory of causation is an account of the meaning of causation. Whether or not A will cause B is a separate matter, and whether or not we can know this, is a separate matter. — Michael

Under a Humean understanding of causation. Not the traditional one. I disagree with Humean causation becues it leads to the problem of induction, and it provides no explanation for why B follows A. It makes everything in the universe contingent.

I'll keep voting and have some victories while you can sit home and let people like me decide your future without opposition. — Philosophim

Your vote doesn't matter. It won't change anything unless you vote in a small enough election where it's possible for one vote to matter. You aren't deciding anything for anyone by voting. The belief that our vote matters is only important on the scale of many voters. Or if you're able to convince enough people to vote a certain way.