disagreeableme.bsky.social
Amateur philosopher, professional software developer, Durham, UK. I enjoy exploring disagreements and trying to understand a variety of views.
829 posts
129 followers
109 following
Active Commenter
comment in response to
post
Specifically oral Chinese. I guess pinyin would do.
comment in response to
post
Have just started watching this for the first time. I did not know that today was Twin Peaks day, so I started a day early.
comment in response to
post
Actually, this is sort of how diffusion models work in AI image generation, now that I think of it.
comment in response to
post
A simple model to illustrate the idea: Imagine you've got an text of thousands of characters which are constantly randomly flipping. All you've got to do to generate a meaningful text is to inhibit most possible flips, and allow only the flips that correspond to the message being generated.
comment in response to
post
I think this is true but not all that profound. I think you can cast any act of generation as arising out of many acts of inhibition.
comment in response to
post
Didn't this basically just happen to Steve Coogan? www.google.com/amp/s/www.bb...
comment in response to
post
Must be because she was being portrayed by a white voice actress, right?
comment in response to
post
That's terminally online.
comment in response to
post
I was with you until the last bit...
comment in response to
post
Nice demonstration of the principle!
comment in response to
post
They seem to be quite capable, more capable than shrimp, anyway, so it's not so easy to rule out LLM consciousness for me (particularly for reasoning models) based on their behaviour. Because, presumably, consciousness pays off in behaviour or it wouldn't have been selected for.
comment in response to
post
OK, so it sounds like lack of freedom is not really evidence of lack of consciousness. As with shrimp, for LLMs we'd need to know what mechanism enables consciousness and see if they have it.
comment in response to
post
Ok, but in a context where we are talking about shrimp consciousness, that seems to be a high bar. I'm not sure shrimp behaviour is any more unpredictable than the behaviour of an LLM. Maybe you would just say shrimp aren't conscious.
comment in response to
post
But it's not a human. It's maybe a very different kind of conscious entity, like a shrimp. Only one that is adapted to mimicking humans. Just because it doesn't mimic humans perfectly doesn't mean it isn't conscious. I'm not saying it is, either, but I think your reasoning is flawed.
comment in response to
post
Ok. This time. Just don't make the same mistake again.
comment in response to
post
You mean it's been the white great shark all along?!?
comment in response to
post
I'm fine with saying they are real also. I should say that rather than matter and energy specifically, I take physicalism to be more about saying that only physical "stuff" exists, whatever that turns out to be. Could be strings, Schrodinger's equation, whatever.
comment in response to
post
Which is why I prefer to describe myself as a naturalist, and not as a physicalist.
comment in response to
post
I think there are those who would say such things are useful fictions, and not actually real. Since we have the words "naturalism" and "physicalism", we may as well put them to use drawing distinctions between people who say such things are useful fictions and those who accept them as fully real.
comment in response to
post
Your naturalism sounds almost more like scientism to me. I tend to agree with you about physicalism, but I prefer to say that naturalism is less restrictive: more or less physicalism about empirical phenomena but openness to moral realism, mathematical realism, etc.
comment in response to
post
I thought this was going to be a minecraft joke...
comment in response to
post
Arguably, Christianity has already been rewritten by millenia of tradition and culture and revision. What Vance is talking about may have little to do with what Christ said, but unfortunately I think he's not wrong in describing the current state of evangelical Christianity in the US.
comment in response to
post
Yes. Several million.
Defining a part isn't going to be much easier. If your point is that defining "simplicity" and "part" is not easy, I agree. If your point is that all rules are very complex, I disagree.
comment in response to
post
The stuff about circuits is more about materialising a rule in a system with few parts. Another aspect of simplicity is that such a system should be able to apply a rule in few discrete steps. There are many kinds of simplicity for different contexts, so it depends on what you're interested in.
comment in response to
post
The second is a more fundamental question about what is more objectively simple, and not just for humans. For the first question, I'd say we want rules humans can quickly and easily apply without external resources. For the second, we want rules that we can materialise in systems with few parts.
comment in response to
post
So, there are two questions we might be interested in here. The first is how do we judge the simplicity of rules when we are specifically discussing number sequences as puzzles for humans to solve. Here, the important thing is that rules be easy for humans to apply.
comment in response to
post
If you need to rely on external resources to apply the rule, then the rule is not simple.
The fact that to build a circuit you need to be really smart doesn't matter either. The simple circuit is just a demonstration of the simplicity of the rule.
comment in response to
post
It sounds like you're just trying to argue for the sake of arguing. The points you are making are, frankly, ridiculous.
I'm saying that we are looking for a simple rule. A simple rule is one you can bake into a simple circuit, or that someone can apply in their head.
comment in response to
post
How about I just publish a web page with an arbitrarily long list of numbers I've chosen pseudorandomly, and then the algorithm is just to read the numbers off the page. That is not a simple rule. That is a rule with a whole load of data baked in, it's just that some of that data is on a website.
comment in response to
post
Forget about all that brain stuff. That's not inherent to the sequence itself. And relying on external resources is also cheating. Think about building a self-contained circuit diagram with no Internet access that has to output these numbers in binary. Your pi-trick is ridiculously complex.
comment in response to
post
If you're looking up a website to find some sequence, then the data that you're looking up is not infinite. If you need all that data to apply the rule, then I'm counting that data as part of the rule.
comment in response to
post
That's pretty desperate. You could look up any sequence that was precomputed. If you're relying on stuff being precomputed, then all this data is included in the rule, making the rule ridiculously complex.
comment in response to
post
But these are puzzles for humans. The rule should be easy for humans to state and follow, and that's the important measure of simplicity.
comment in response to
post
Where you have a point is that any sequence that relies on hunan domain knowledge (like how many letters are in the English names of numbers) outside of maths itself is cheating a bit by containing a bunch of arbitrary stuff that makes the rule inelegant.
comment in response to
post
Well, you have a point, but that's not literally true. The sequence 1,2,3... is a sequence I've just thought of but does not require a lot of computation to reproduce with, say, a Python program.
comment in response to
post
Good one. Let's say that rules that require a lot of computation do not count as simple.
comment in response to
post
It's fine! I just disagree with it. I think talk of qualia and "there-is-something-it-is-like-to-be" isn't really tracking anything. I don't think MaryGPT has qualia and I don't think we have qualia. I don't think MaryGPT is conscious because it's too different from human info processing.
comment in response to
post
Doesn't seem so simple. It's got a lot of free parameters. Each number you provide explicitly is making the rule less parsimonious.
comment in response to
post
If you can come up with one simple rule that can generate the given sequence (deterministically), then that rule determines the next. A good problem won't have more than one such maximally simple rule.
comment in response to
post
Hmm, OK, so I guess it has to be a predictive rule that would allow you to give the next number. Simplicity is somewhat subjective, but typically, in these problems, there is only one standout answer.
comment in response to
post
It used to annoy me when I was a kid, because the answer could be anything. But now I reframe it as: what's the simplest rule you can think of that can account for this sequence? Then it's not quite as arbitrary.
comment in response to
post
5, 5, 4, 3, 6...
comment in response to
post
And I don't think that's really what we mean by "consciousness". I think we mean something that has experience a little like ours. And that requires processing information like we do.
comment in response to
post
I don't think there are any necessary or sufficient conditions here, just family resemblances. And the way the US processes information in the real world is nit really all that like how a brain does. In principle, a thought experiment like China Brain is possible, in my view.
comment in response to
post
I think that's what it is, yes. But that doesn't mean dogs or other animals are not conscious. I think dogs process info much like humans do, even though they lack a lot of the cognitive abilities.
comment in response to
post
Ok, so my point is that there are a number of different possible responses to ES's challenge. My response is different from yours. This suggests to me that he's not commiting some egregious error. Even if you think you have the answer, that's great, but the answer is not an obvious one.
comment in response to
post
Ok, and to the extent that it really does make sense to treat them as an agent, to take the intentional stance to them, then it is no more absurd to postulate that they are conscious than that the US is conscious. Eric would agree, I'm sure. There's nothing special about nations in this regard.
comment in response to
post
You asking me or Eric?
Me: No, because the group of left-handed NBA fans to not, collectively, process information in a human-like manner.
Eric: No, because the group of left-handed NBA fans do not collectively behave as an agent with goals and intentions.