Artificial Intelligence Is Stupid and Causal Reasoning Will Not Fix It

It is, but I think its speed comes from the fact that it doesn't have to go through any discrete states during the computation. There's a start state where the weights are hung off the chain, then the end state where it settles into a new shape, but what is between? An infinite number of states? You can perturb it, even, while it's settling, and it will still not change its end state, which is not something you can do to any kind of machine. The computation is also "free" as in gravity. I don't know how we could build a machine to do that and I don't think we can describe it as a Turing machine.

Btw, what makes it a computational device is that the shape of the catenary matches the shape of a load-bearing arch. It's one physical system calculating another. So it's not just using the physics of one system to calculate itself.

But you are right. The universe calculates its own state faster than any computer running in-universe could. Well, why? That's just one thing we don't fully understand. Maybe thinking of the universe as a computer is applying the wrong analogy, and maybe we do the same thing with minds, as you say about electricity etc.

I can't say I understand Penrose's arguments very well. I'm not a phycisist. Turing equivalence exists, of course. But there is something to the fact that I don't get stuck in a loop if I think that "this sentence is false".

So the proposed arrogance is the 'metaphysical arrogance' that attributes such a profound capability with a few kilograms of organic matter. Interestingly in Qur'an, this sin of 'arrogance', and, associating 'sentient worth' to its material composition (the famous statement of satan that it is "made of fiery matter while Adam is made of clay") is a declaration of satan. This then circles back to other comments in this thread questioning the Organic Matter Mind (clay centric) chauvinism.

> What are the proofs that someone unable to perceive anything from the physical realm can be conscious?

John Lilly famously did his isolation tank thing to get to the bottom of that, in this diminished (but not totally occultated) sensory environment. He wrote books about the experience. I personally find the idea that mere structure gives rise to mind to be 'wishful' (if not arrogant) thinking.

So if you want to know and you are unwilling to settle for here-say and logical deductions, you will have to dive deep!!

Mind you I am not advocating "my" tradition, or telling you to buy my book or whatever nonesense. I beg you on my knees to sit and close your eyes and go inside!! It is horrible in the beginning because of ALL THE TRASH that we accumulate in our minds. You have no peace, no tranquility, no joy and no bliss. But for this much of a leap of faith I beg you pardon: IT IS THERE! There is absolute tranquility that leads to bliss to be attained as a _sideeffect_ of your quest for truth! Truth is not what I tell or anyone else! Truth is what you experience. The only question is how deep that experience is.

All the mystic traditions have thaught the same. NO REASON to believe, but hopefully enough reason to risk to try it!!

(And besides that, you will get all the benefits with that come with meditation and cleaning up the mess of the mind, which have already been proven by science)

If an assertion sounds like it makes sense, that doesn't mean it's not arrogant to make. Another point is that you supposing that since it "makes sense", that's enough justification to pass the burden proving-it-isn't-plausible onto someone who disagrees. In effect it comes off like someone taking one person for their word while asking for documentation and proof from another. Consciousness is an elusive thing to define much less to start making super strong claims about at this point lest you reveal some internal prejudices.

A person who has had their heart replaced with a pig's heart does not suddenly lack consciousness or have the consciousness of a pig. Extremely short people, such as Afshin Ghaderzadeh 2'1", don't have less consciousness than extremely tall people, such as Robert Pershing Wadlow 8'11".

Stating that consciousness is not an emergent property of the whole body should be a starting point, after which further research and discussion can occur.

Other interesting thought experiment: if you developed as a baby with the complete absence of senses. No sight, touch, smell, etc. would you ever develop consciousness? I suspect not. I believe our inner subjectivity is inexorably tied to our senses.

Of course there are still criticism, this is how scientific research works. Criticism does not mean disproven, mind you! It is a "new" theory with a lot of promise, that needs a lot more work

Extraordinary claims require extraordinary evidence, and absence of evidence is not evidence of absence.

To your point 2, this directly contradicts your point 1, and is, what I would call, a "dualist" philosophy. This is a basic statement that thinking, the way we do it, is fundamentally not mechanical and therefore not describable by our laws of physics. That is, however we think is literally not of this universe. The "dualism" comes in by naming the other "plane of existence" (or whatever you want to call it) as the companion universe or reality to ours where our intelligence effectively resides.

I think Turing is a little maligned here. His test was meant to strip away all the red herrings that one might use and focus on what (I'm imagining) he thought was the most fundamental of all human traits, the ability to reason and to use language. The Turing test is a functional one. If something talks like a human, reasons like a human, acts like a human, then how would you tell the difference between the "real" human and the "simulated" human? By creating a functional test, it gets at the core of the issue. If you can't come up with a test to differentiate one from the other, the question of what's "really" human becomes academic.

So, to, the focus on Turing machines. The specifics are a little byzantine, nowadays, but the point is that there's an "equivalence" of machines, where one can be used to simulate another. The point about being Turing machine equivalent is the mechanization part, not the specifics of tape reading and what the state machine looks like, so much.

In terms of proof, I would say this: If someone were to predict the progress of machine learning and artificial intelligence as it progresses towards becoming feature parity with the human brain, one might well predict all of the milestones that have been achieved. For example, beating the best human at checkers/chess/go, being able to do speech to text, being able to identify people by faces, being able to replicate someones voice/gait/face, being able draw novel art, being able to hold on a conversation, etc.

In my opinion, the human brain is not monolithic and is compromised of many sub systems interacting with each other. That being the case, I would expect some type of "general intelligence" to come by once all the sub systems have become modular and easy to use, with some kind of "operating system" bringing it together to be able to have one part communicate with another and be able to make broader scoped decisions based on input from each of the subsystems.

This is a long way of saying I believe there is quite a bit of evidence that we're getting closer to "thinking machines". They may not work exactly how our brains work but that's kind of beside the point, they're functionally doing what we would expect them to do on the path towards human brain feature parity.

- 'A' is a computing entity and an organism: it takes information in, it reacts. Its genes use those two functions to ensure survival. Note that A doesn't need to be conscious.

- As environmental complexity raises, A must collaborate with other members of its own species (most likely, the same genes control the hardware) or of other species. Let's say B.

- Evolution gives A the capability of partially simulating B ("other"), so that they can negotiate. Evolution also gives them the capability to communicate. Communication happens between members of the same species because they share the same set of genes. If B survives, the genes of A survive, even if A itself doesn't (abnegation).

- The capabilities are somewhat loose. For example, there is not a restriction that A must be different from B. If A can simulate B/other, it an also simulate with B/other=A. The trick is very useful to strategize interaction ("plan"), and to cope with bigger groups.

That process where our brains diverge and simulate an interior dialogue is what we perceive as consciousness. For example, as I write and edit this post, I'm actively imagining ('simulating') how peers from my species will interpret it, and my sentences are changing so that they are decoded by the path of least resistance in the reader, thus advancing my goals (actually my genes').

So, consciousness is an illusion that we can't escape in speech[^1], discourse or logic. It is what we are. And it is tremendously useful, because it allows us to "edit our future" to a great extent, with the implication that by doing so, we increase the chances of own survival or, if that's not possible, of our genes' survival.

Now, that component is missing from our current AI iteration. Put simply, our artilects don't run continuously and don't have genes commanding them. They don't have an endocrine and nervous system constantly producing overriding signals ("hunger", "pain", "lust", "do so that the group survives") that must be rationalized a-posteriori to the point of forming entire cultures. The thing is, we could engineer entirely artificial replacements of those things, and with a bit of fussing, we would replicate the illusion of consciousness. But the idea of doing that and also binding our creations to do useful work is simply repulsive to me (edit: my genes are providing an overriding signal saying that, creating a being with the potential to inflict damage on me and my group (read: them the genes), is something that only deleterious members of the species would do, and that at the very least, dissuasive action should be taken).

[^1]: We can shut it down however, it's not inherent. We routinely do when we sleep, and sometimes when we do something that requires more of our lower brain, for example doing intense physical exercise or having sex.

Machine Learning (ML) is a catch-all term (not in of itself a technology) for any approach where systems are designed to learn from data. This includes techniques such as random forests and SVMs, as well as neural nets.

There's really no fundamental relationship between AI and ML. Neither one is a subset of the other as this is an apples and oranges comparison - one is a goal, and the other is an approach). That said, all recent progress towards AI has been achieved by using ML, although not all uses of ML can really be regarded as AI.

Hope that helps define terms.

I'd describe it still a language model, but just one with "filtered" output.

I'm not sure if ChatGPT has been documented/described, but it's very similar to OpenAI's InstructGPT which they have described, and which they still refer to as a language model.

> We’ve trained language models that are much better at following user intentions than GPT-3 while also making them more truthful and less toxic, using techniques developed through our alignment research. These InstructGPT models, which are trained with humans in the loop, are now deployed as the default language models on our API.

https://openai.com/blog/instruction-following/

I think it’s not a very good metric.

https://i.imgur.com/im9EquK.png

Perhaps it's magic trick that a language model can do math, but it's quite impressive that it gets very close to the desired answer.

They might make really bad decisions in edge cases(Which will get less and less common as more cars get smart), but they might make up for it with perfect behavior in ordinary circumstances.

They will never prioritize convenience or speed or avoiding angering other people over safety. They'll do the safe thing even if no human driver could maintain that level of paranoia at all times.

"but can it be done is no longer in question."

You are right - it doesn't seem like it can be done at all, not in the way that most people imagine "self driving" - where you hop in your car no matter the weather, tell it where to go and it just takes you there while you sleep, and it works in every country in the world on every road like any human can drive.That doesn't seem possible this century, if ever.

The question that remains is whether it *can be done* in an economical manner that is at least as reliable as a human driver. Anything less is a failure in terms of money and lives impacted or lost.

But I've got no idea how practically usable such an approach would be.

I also have tesla’s fsd beta on my car, and it does a marvelous job not hitting people. If anything it’s way too cautious around pedestrians.

It could be argued that there are already ML systems that do what you describe (i.e. make instant effective judgements/decisions about something it hasn't been exposed to during training), at least for some contexts. For more information: https://en.wikipedia.org/wiki/Zero-shot_learning

Current methods are far from perfect, but not for the reasons Marcus believes. Consider that we /should/ be able to derive hypotheses for better system behavior from a given critique, test those proposed system improvements, and then (if the hypothesis is correct) see some improvement. Marcus' crew hasn't produced any useful improvements; this is either because they are not technically competent or their hypotheses are wrong. Given that people have been hammering at these systems for ~10 years now, I doubt the failure to build better logic-based systems (or hybrid systems) is for lack of trying.

Meanwhile, the non-Marcus groups have been improving systems by leaps and bounds, through exactly the kind of iterative, hypothesis-driven improvement I described above. Certainly nothing is perfect, but we can see real progress happening.

(Finally; Tesla FSD is a strawman. It's Waymo you need to argue with, and they're doing great, AFAICT.)

Given any well-defined (Kurzweil/whatever) variant of the test, we can obviously test any entity and see if it passes. However, my assertion is that passing the test is "meaningless" in sense of saying anything profound about the entity that passed.

So, there is no either/or consideration here - we can discuss an entity passing the test (benchmark) regardless of what importance we attach to the result.

My point was simply that fooling people into believing something isn't the same as that something actually being true. Additionally, as far as any "intelligence" (and/or "consciousness") test goes, IMO it's even more meaningless if we attempt to test for these before defining precisely what we mean by these terms, since otherwise passing is a subjective measure ("intelligent in the opinion of the test supervisor").

I'm OK with the notion of using a black box test to gauge intelligence (since I regard it as behavioral quality), but we have to be careful when comparing to a human (as the Turing test does) since we could allow the bar to be arbitrarily low by not demanding the performance of a fully intelligent and unhampered human (can ChatGPT convince someone it's an idiot with OCD stuck in a box looking up answers in a book it doesn't understand ... maybe. can it convince me it's an averagely intelligent student able to perform experiments and report on the outcome - not even remotely close).