This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
They said language and meaning stood outside of mathematics, it was some uniquely human faculty, beyond cold logic.
But then we added more layers of matrix multiplication. Language and meaning turn out to be not so far removed from mathematical principles after all.
None of that impinges on the mountain of evidence that the human mind cannot be manipulated deterministically, whereas our machines generally can be, nor on why this difference should be disregarded. You are claiming that the brain is essentially a computer, "albeit with great complexity", but in fact all the things that make a computer a computer cannot be found in any of the parts of a human brain we directly observe, and can only be supposed, evidence free, to be hiding in the "great complexity" part. This, after several decades of scientists claiming to be able to both locate and interact with them and uniformly failing.
AIs are extremely amazing, but from working with them myself it seems pretty clear that there's not actually anything resembling a mind in there, at least in the versions I've used the way I've used them. I think it's entirely possible that additional matrix multiplication might change that in the future, but right now, it's doesn't appear to me to be happening. As mentioned elsewhere in the thread, the bot appears to me to be channeling intelligence, not possessing it. It doesn't seem to me that the difference is subtle, though I am again open to being persuaded that my technique is bad or the models I'm using aren't the good ones.
Can you clarify what you mean by this? What qualifies as deterministic in your view? To me, things like alcohol or lack of oxygen impacting cognition in very predictable ways seem like good evidence that there are ways to alter mental states deterministically. Certainly not in a particularly precise way at current levels of biological knowledge, but clearly not random or immune to structured intervention either. For example, my uncle very consistently goes from sweet and caring old man to raging asshole after a couple of drinks, and I feel it's pretty intuitive to understand this in a model where that is a straightforward deterministic effect of the alcohol on his brain chemistry.
At least weak versions of mind reading or mind control, essentially. Science as a social institution spent the first half of the twentieth century claiming not only that these were possible, but that it had practical methods for implementing them.
Are you aware of a reliable cure for alcoholism?
I'm aware of the general vibe you're pointing at, but I've honestly never come across a strong claim by someone I'd consider a serious scientist (i.e. someone who has made a meaningful contribution in some scientific field) that resembles anything akin to mind control or mind reading, that's more or less the domain of speculative fiction authors.
What is commonly argued, and what I personally also find defensible given what we know about material reality, is that the human brain, which for the purposes of this argument is assumed to be equivalent to the mind in terms of functional behavior apparent to outside observers, is in principle a deterministic material thing and thus predictable just like any other material system is, again in principle. Noticeably, that claim doesn't entail that we will ever find sufficient models and theories to achieve that predictability, nor that it's even possible in the first place. The obvious parallel here is the weather or n-body systems. The constituent interactions that make up these things are known with high certainty to be deterministic, but the overall system in each case is so chaotic that predicting it with meaningful precision is impossible for large timescales, because even in the best of models the errors keep adding up.
I'm hearing good things about Ozempic, though it's not as effective as it is with hunger. That's probably a much stronger example I should have gone with in the first place.
John Watson
B.F. Skinner
At least arguably Freud, depending on how much weight you're willing to give to his motte-and-bailey caveats.
At least arguably Dewey and many of his disciples in the blank-slate educators camp.
Marx, notably not a scientist, but a whole lot of scientists don't seem to have been able to tell.
I have zero argument with an axiom that the human mind is a materialistic machine, so long as it's recognized that this is, in fact, being used as an axiom, not as something that's been empirically proven. My objection is to the often-conflated claim that the brain has been empirically proven to be a machine. I object to this because the great majority of useful things that we can do to and with machines that make the label such a useful one, we can't actually do to or with the brain.
I'm on Ozempic, and it's not exactly a silver bullet for hunger either. But in the case of both alcohol and Ozempic, we're still a long, long way from mechanistic interactions, right?
I concede that these are notable people in their fields and that they made all sorts of dumb predictions. But I hope you'll forgive me if I engage in some "No true Scotsman"-ing: none of these are who I thought of when I wrote "scientists" (although I do admit in common usage they also fall under that term). All of these people constructed elaborate theories without much empirical feedback or validation, I was thinking more of people in the hard sciences producing testable results than then tried extending out into making claims about the brain.
It depends on what you mean with empirically proven and how much weight you put on the word "machine". Again, I fully concede that a predictive theory of the brain as a whole is nowhere to be found in current science. But all isolated parts we've ever looked at look like other matter, which at our scales behaves according to deterministic rules. And again, using the weather or n-body examples, we already know that deterministic systems can quickly escape predictability. The sort-of-empirical argument then is that the stuff the brain is made out of is deterministic, the small components and subsystems are deterministic, and since there doesn't seem to be a good argument for where non-deterministic behavior would come from, it stands to reason that the resulting system is deterministic, too, even if it is so unfathomably complex that a full accurate description might never exist.
I'm not sure I agree. It depends on the scale of causal and mechanistic resolution that you are willing to accept as a sufficient base level. Newton famously caused a lot of consternation among his scientist colleagues because his model replaced a universe of metaphoric pistons and levers acting directly on each other with spooky action at a distance.
For Ozempic, we don't know exactly how the concrete interactions work out in altering specific emotional states, but we do know that it "attacks" the parts of the brain responsible for hunger and reward signals (this we know from simple observation), and how that affects downstream calorie consumption. That's not an exhaustive model, sure, but the statistical evidence from placebo studies and from normal usage across millions of people is IMO more than enough to be able to positively claim that this is a material intervention that results in a broadly predictable outcome on the brain.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
A guy used TNT back in the day, blew a hole straight through his head and that certainly affected his personality, he got all aggressive and dour. Neuralink. Lobotomies. So much for uniform failure. It's just that the interaction is fairly primitive compared to what we can do with AIs, for a mix of technical and ethical reasons.
We have a map of what the brain does, we've isolated what parts do language or other roles and can turn parts off and on. You underestimate what progress has been made in neuroscience: https://www.sciencedaily.com/releases/2015/10/151014084955.htm
What is the distinction between channelling intelligence and possessing it? AIs can reason, plan, problem-solve. That's intelligence. What specifically is missing? I too use Opus and Fable, I'm satisfied that they're intelligent. They are not like humans in all respects of course but they certainly possess intellect.
Black powder, actually, and last time I read up on him, I recall a fair amount of evidence for replication crisis issues. That is, there's evidence the scientists studying him had an agenda, and shaped their studies and the reporting of those studies to fit that agenda, rather than impartially reporting facts. which actually feeds directly into:
Neither impinge on the objection I'm raising, though the actual history of Lobotomies is actually a really good example of the history I'm pointing to. Psychology and Psychiatry and the medical field generally repeatedly claimed they could deliver practical mind reading or mind control, gained massive sociopolitical power from these claims, and then failed to deliver them while gaslighting the public into a messy, ad-hoc combination of "we delivered them" and "we never promised to deliver them". Massive parts of history make a lot more sense if you understand this, because this particular scam was wildly impactful on the political history of the Modern era. If you want a bit of the flavor, read old Sci-Fi, it's absolutely riddled with the resulting tropes. If you dig into the political and institutional histories, you'll find the same thing, because the Sci-Fi was extrapolating from the understanding of science promulgated by the actual scientists. Lobotomies were established as a legitimate technique because the scientists claimed that psycho-surgery was a real thing they could do, and fell when it became unavoidably evident that those claims were categorically false.
I don't think I do. Mind Reading and Mind Control were the table stakes proposed by the Humanist contingent in the turn of the last century, and it does not appear to me that we are any closer to these now than we were then. If the brain is a computer, we are researching how to plug in different keyboards, monitors, mice and headphones, and at the bleeding edge we're figuring out how to bypass damaged USB ports by soldering wires directly to the board, while the modernist claim you seem to be repeating is that we can edit the source code. The distinction matters. In this too, I'm open to being proven wrong: maybe next year we'll actually have mind reading and mind control. But I observe that one of the worst and most common failures of the Scientific Materialist mindset is the propensity to assume things are empirically proven when they haven't. And again, mind reading and mind control is both what we should expect if the brain really is like a computer, and something science as an institution started claiming it could deliver something like a hundred years ago.
I'm making mathematical models of tank guns. The AI is amazingly knowledgeable about physics, engineering, math generally, military history, and on and on. I say I want to figure out a general model for gun weight based on peak pressure generated by the propellent and the shell. It can do this! It calculates hoop-stress for steel, calculates the pressure curve...
...and then it explains to me that shortening a gun's barrel increases the peak pressure its shell generates. Only, that's not true. Hold the shell weight and propellent charge constant, and barrel length cannot increase peak pressure, only reduce it. I'm not particularly proficient in physics or mathematics or engineering, but even I know that. It's simple logic if you understand even the basics of what these words in combination are describing.
It "knows" a vast amount, and it presents its "knowledge" in what appears to be a very human way from the perspective of the user, but the knowledge is not integrated or manipulated in the way I would expect from a human. The gaps in knowledge and certainty appear in places I've never seen in humans, and working around those gaps involves a different process than I've ever used with humans. The AI doesn't actually understand anything about hoop stress or the physical properties of steel or the internal ballistics of tank cannons, or possibly anything at all. It's not an actor in its own right, the way we commonly think of them. It's a lens, some sort of high-dimensional refractor that concentrates diffuse, disparate elements into a specific focal point, but fundamentally dependent on the physical positioning of those elements and of the lens itself for the content of that focal point.
...And again, maybe this is me just using it wrong. If so, I'm very interested in how to use it better! I'm only a bit over a month into my career as a serious AI prompter, so I claim no deep insight into the actual frontier capabilities of all AI or even the models I'm using. I'm just describing what I'm personally seeing. Further, even this is fundamentally amazing! We've legitimately created a repeatable, mass-production terminal for coding in Grad Student! That by itself with no further improvement is probably the biggest technological breakthrough of the last century! But while it presents itself as a person, it pretty clearly isn't a person, and the illusion only holds even briefly at the focal point itself. sidestep, and it fuzzes out and breaks apart.
OK, so they oversold what could be done with crude brain surgeries. But wireheading definitely works, there was that woman who rubbed her thumbs raw against the button. How is that not direct interaction with the brain? You switched from 'uniform failure' to 'they were overselling the practical benefits of the technologies that were widely deployed, even though this was due to ethical rather than purely technological reasons.'
Lobotomies also work, it's just that they aren't really helpful.
I never claimed that we could directly edit the source code of the brain right now (noting that source code is not quite analogous). But eventually something like that is going to be possible. I don't understand how you could possibly disagree with its plausibility. We could for instance add or remove certain genes that are strongly linked with aggression in an embryo for example. Add the warrior gene, get a more aggressive child. The source code is genes, we can edit genes. We don't quite understand what they all do and how they work together and it's easier to wreck than improve but we could definitely do these things.
Editing the brain 'in flight' would be rather harder but it's a physical object, it could be altered with sufficiently precise electrodes or other manipulations. Brain surgery is a real thing. This would just be an extension of brain surgery. Why are you so skeptical of this?
Furthermore, I linked that old paper about disabling parts of the brain with magnets. That is a kind of mind control, just not very useful. Cheap, precise, non-invasive, long-lasting, permitted by the ethics board... That's the missing part, there's no actual impossibility.
OK, it made a mistake? Shortening the barrel increases muzzle pressure but not peak pressure.
I use AIs a lot. They make mistakes in my projects, I suspect they might not be smart enough to do some of the most complicated tasks proficiently. Why should that mean they were not intelligent? Many humans also make mistakes. Give a basic, basic statistics/probability/science test to your national representatives, see how they perform. Half the British Parliament couldn't tell you the chance of two coins coming up heads. Few humans can make a 4X AI actually smart, it's very rare to see.
If AI weren't intelligent it wouldn't be able to get anything correct at all about hoop strength or do any maths or coding. If it weren't intelligent you wouldn't be using it at all. You wouldn't, couldn't use GPT-2 for your project because it would just babble nonsense at you. That's a true lack of intelligence, you'd need to squint to find its intellect. The original GPT-4 was a bit like this too, context limit too short for much serious work past a certain level of complexity (I tried), too many hallucinations. Sonnet 3.5 also had its own flaws, it got incoherent past a certain point and I'd just have to say 'ok it still doesn't work, come at this from another angle' or 'think outside the box' and it was painful getting anywhere.
But their inability then didn't mean they weren't intelligent, just that their intelligence was weaker.
Just because it doesn't fail in the same ways that humans fail, doesn't mean it's not intelligent at all. I don't claim that they are human, there are many aspects where they're clearly not. But some important aspects where they are like people and could reasonably be expected to have other human properties.
Again: I am absolutely not dogmatic about the following, just groping for description of my experience.
They seem smart enough to do almost anything, but only when actual crystalized smartness is arranged in a properly-aligned constellation around them such that the focal geometry works properly. As soon as the alignment is broken, they get real stupid real fast. From this I infer that they are not themselves intelligent; the constellation is where the intelligence is, they're only focusing it, or channeling it, or pumping it.
Actual intelligence would actively align itself, would internalize the constellation in a cohesive way. And the AI seems to do this, just to a very limited degree? Maybe, even probably, it'll do it better as the tech matures? Certainly I can imagine an AI actually built for the work I'm trying to do with it, which has in fact internalized these specific constellations to a degree indistinguishable from a human expert, and we just aren't there yet for 1900s-era tank cannon design simulations.
In any case, it's not that it's stupid; as you say, humans are stupid too. It's the way it's stupid, how stupidity spikes massively and unpredictably. The "channeling intelligence" model would seem to imply that this is going to be a chronic problem in the long term, but I'm just some dude bullshitting on the internet about my bottom-tier subscription shenanigans. I'm pretty excited to see how it goes.
Is a for loop intelligent, in any meaningful sense? are intelligence and automation synonyms, and it really is just a matter of scale? I would bet no, at least in terms of practical scale, but I'd readily admit that my position is largely held by inertia and that I'm standing to be corrected here. There's very plausibly going to come a point in the next year or two when the focal geometry auto-focuses in real-time and you can't find the seams any more, and the question becomes something you should probably just ask the machine itself and accept that the answer is the best you're going to get.
on to the more contentious part:
This statement is fundamentally wrong. Lobotomies do something; they aren't a sugar pill. They absolutely do not work, as in do the thing their developers and promoters claimed they did to get them accepted as a valid medical treatment. The difference, I think, ought to be instructive in a context where I am pointing out that your claims vastly exceed the empirical evidence.
And the field and the public at large overbought. And this pattern repeats itself everywhere throughout the history of psychology as a "science". You yourself are, I argue, overbuying the claim that "the brain is a computer". If the brain is a computer, it is a computer that we do not know how to do any computer-specific things with or to, and that is an observable fact.
We have heroin. how is heroin addiction not direct interaction with the brain? Show me how wireheading can be used to actually control people as one controls a machine, in any one of the millions and millions of ways that being able to control a specific person as one does a machine would be immediately and obviously useful. This is my understanding of your claim, and it is the claim I am disagreeing with. You said:
My understanding is that you are claiming the brain is a machine. I am asking you to show me a way that we can do machine shit with it. Smashing a machine with a baseball bat is not using it as a machine. Melting it into a brightly glowing puddle is not using it as a machine either. sitting it on a turntable and spinning it also is not using it as a machine. If the brain is a computer, show me code executing arbitrary operations, the way computers do. If the brain is a can opener, show me how you open a can of beans with it. And show me how you do these things against the will of the person currently occupying that brain, which is the obvious contrary hypothesis that is supposed to be refuted here.
It might be. I'll believe it when I see it. Until I can see it, I'm going to point out that this is a faith-based claim.
The mechanics of brain surgery are well understood in most cases, as they are an extension of the mechanics of regular surgery: stuff has been broken apart, you stitch it back together so it can heal. As noted above, this has not resulted in the development of mind surgery, despite a bunch of prominent scientists claiming it absolutely had, and then brain-damaging a bunch of people in a wildly irresponsible manner before common knowledge accumulated that they had lied.
Because vast evil has been resulted from people being insufficiently skeptical of "scientific" "materialism" that was neither scientific nor materialist, and lobotomies are the very least of it. Similar claims about this specific subject have caused significant individual and collective harms over the course of the last century, because trusted authorities asserted them in a confident voice and sufficient skepticism could not be mustered to dissuade them. A huge portion of the culture war has been directly driven by these sorts of claims.
You are presenting this:
...as a scientific, materialist claim. It is not scientific, and it is not materialist. That doesn't mean it is wrong, it means it can't actually be tested via the scientific method and the results observed. You think it ought to be true, that it is highly likely to be true, by inference from other scientific, materialist claims that have been verified, and that is well enough. But there not only is not solid, direct evidence that this is true, but there is solid evidence that it is not true: all the previous theories attempting to act productively on it that demonstrably failed when implemented, and the subsequent retreat into Determinism of the gaps.
I'll believe it when I see it. I used to be 100% on board with the idea that the brain was a computer and all that this would imply. But a big part of what made me on board with it was that I wanted to believe it was true, because I'd been sold all the shiny ideas and cool things we could do if it were true. Then I got the other half of the pitch, and my conclusion is two-fold: firstly, we do not actually have a solid basis to claim that the brain is a computer, and secondly, if the brain is in fact a computer, I very much hope I'm safely dead before they figure out mature brain-computer-science.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link