Dennis Hackethal’s Blog
My blog about philosophy, coding, and anything else that interests me.
Tweets
An archive of my tweets and retweets through . They may be formatted slightly differently than on Twitter. API access has since gotten prohibitively expensive – I don't know whether or when I'll be able to update this archive.
But in case I will, you can subscribe via RSS – without a Twitter account. Rationale
The content of an explanation is a statement about what is out there in reality, how it works, and why.
Starting with observations leads to an infinite regress you fail to address. You seem to willingly ignore it, idk why.
You're displaying irrational discussion methodology by repeating your point over and over without addressing criticisms.
Again, they might well have an explanation of what the dog's nose looks like, yes. (That they wouldn't know what it would feel like seeing a dog's nose doesn't change this.)
It's not so much about thought. It's about knowledge: solutions to problems.
"Everyone knows that to do great work you need both natural ability and determination."
What do you mean by "natural ability"? Talent?
This isn't a productive way to think about knowledge, though. As I said, knowledge creation begins with problems, not observations.
A creative mind can - in principle, though extremely unlikely - happen to come up with a theory describing dog's noses without having ever seen them. What it would feel like seeing a dog's nose is a different matter (quale).
That's the right way to think about our role in the cosmos.
And hopefully, re the original post, we as individuals are going to find a way to be here for much longer than tens of thousands of days. twitter.com/ChipkinLogan/s…
No. You're putting words in my mouth. I have already agreed that blind people would not know what red feels like. I have also explained that this concerns qualia, not knowledge generally. This is part of the root of your mistake: you keep mixing up qualia with knowledge.
You keep describing knowledge as if it was already present. It isn't. It needs to be created through conjecture. Such as how to observe the delta between blue and red.
Explanations in terms of neural structures/hardware not fruitful. Intelligence is software, not hardware.
"Change is x to y." Yes. You know this from theory, not observation.
No. You know about change because you conjecture a theory that explains why things you observe are not always the same. Has nothing to do with neurons or underlying hardware.
Moral knowledge consists of components that cannot be observed. So where does it come from?
You know what change is from conjecture, not from observation.
So I ask you again where knowledge comes from that couldn’t possibly be observed, which you still haven’t answered.
Change is not observed either, btw.
Yes. It can wonder if there is anything other than darkness, what to do with its life, what to want, if this state will ever change, etc. And it can - tentatively, fallibly - answer these questions through conjecture.
None. But that’s fine. To create knowledge, we start with problems not observations.
LOL. If you just hide behind IP once the going gets tough, you haven't explained anything.
And how did physicists know how to combine these things to get to the concept of fusion?
And what observed concepts do I use to create things like moral knowledge, which does not consist of any observations at all?
So what are the building blocks of fusion that physicists observed?
Indeed. And, we know all kinds of things we could not possibly observe: we know how hot the insides of stars are, how massive black holes are... and we have moral knowledge, aesthetic knowledge, etc, which cannot be observed either, nor can their building blocks.
Nor have you explained what your AGI design is, or why the Popperian notion of knowledge is insufficient, which is what you set out to do.
You clearly are not familiar with the Popperian notion of knowledge, or you would have known that it's explanations.
If we make our own knowledge all the time without observation, why do you think that observation is so important?
Also, you haven't explained yet what knowledge is, only vaguely that we build it from observed "components".
The theory explains why empiricism (which is what you’re espousing) is mistaken, why neuroscience and other narrow AI is heading down blind alleys, it explains what knowledge is and (roughly) how to create it. What can yours do? Recognize shapes.
People create knowledge by sitting in a dark room all the time. Close your eyes and try solve problems. The solutions you create are explanations you create.
They also contain approximations to explanations of how to make wings, build brains, and may even contain approximations to the laws of physics.
Genes contain explanations of how to spread themselves through the population.
That's because you have the wrong idea about what knowledge is. Knowledge is explanations - statements that are adapted to solving problems. Those don't come to us through the senses. We have to create them ourselves through an evolutionary algorithm.
No. It is evolution that creates knowledge generally, both genetic and human.
BTW, these things need to be explained on the applicable level of emergence. Intelligence is a property of software, not of some underlying physical structures in the brain.
But knowledge does exist in the brain in the form of the structure of cortical columns; plus mechanisms of how to associate them given certain sense data, etc.
Knowledge - of any kind - is created by evolution: variation alternating with criticism.
Indeed. So biological evolution creates the genetic knowledge of how to do the wiring in such a way.
If biological evolution explains the origin of genetic knowledge, why should an evolutionary algorithm running on the brain not explain the origin of human knowledge?
In principle, but I agreed that it would be unlikely to happen.
Anyway, okay, so let's say those components are stored in cortical columns. Where does the knowledge of how to store stimuli in cortical columns come from?
I didn’t say blind people can see color.
Okay, so let’s say those components come to us through the senses. How?
Faith has nothing to do with it.
I had already granted that it would be extremely unlikely to think of dogs in that situation, but that's incidental.
You seem to claim that knowledge comes to us through the senses. Yes?
- For example, that it has legs, a nose, etc.
- Because in the sense you mean, red is a quale. We don't understand qualia, so I don't know.
I'm familiar with the argument you're making. Brain in a vat. I could think all kinds of things about dogs I want. Anything thinkable I can think.
If you were to say that I cannot predict the quale of seeing a dog, then I'm with you.
I'm in principle capable to have any thoughts about cars and dogs I like. I would be extraordinarily unlikely to have them, though.
@HeuristicWorld
As a preprogrammed emotional response, sure, but no associated quale of suffering.
@HeuristicWorld
I think there is no animal behavior that cannot be explained in terms of genetically given algorithms that just need to be executed; that leaves no room for creativity.
If consciousness, suffering etc arise out of creativity and animals are not creative, they are not conscious.
As if the bastardization of the term "AI" had not gone far enough, folks are now starting to bastardize the term "AGI".
If this spreads, we will soon need to find a new term to talk about the real thing again. twitter.com/markcannon5/st…
You're building a shape recognition algorithm. That's not AGI, even if it can recognize all kinds of shapes.
Well, we have a good explanations of why it would be a bad idea to unilaterally disarm.
Using justificationism against justificationists is an interesting approach, however. But does that not change your yardstick for what you consider real? I.e. a good explanation vs "proof"?
Burden of proof is justificationist. My link refutes Mark's points. He now needs to either explain why it does not, or refute the link.
@sciencemagazine @ScienceCareers
This isn't science. This is social justice nonsense.
"Knowledge of how to use the senses is neither encoded genetically, nor can it possibly come from the senses. So where does it come from?"
About the folly of empiricism and the recovery from blindness:
E.g. stoning a woman to death because she dared take off her hijab makes me cringe; not because the perpetrators are ignorant - we are all infinitely (though unequally) ignorant - but because the ideas behind it are deeply false (and in this case, deadly).
Ah, it's not the ignorance per se that's cringeworthy (by some criterion): it's the badly mistaken idea. Some ideas are worse than others in terms of the damage they cause.
Some ideas are cringeworthy, but we may all have different thresholds.
People like Feynman, so they rush to his defense. Ad hominem. And unnecessary: he would have liked to know he was wrong so that he may improve.
Wheeler (both Feynman's thesis advisor and David's boss) knew Popperian epistemology well and may have introduced Feynman to Popper, but it isn't clear.
I know that you know who Wheeler was, but including it here in case others read it and aren't aware of the connections.
Yeah. Quote from David:
[...] I happened to mention Popper in the one conversation I had with Feynman, sometime in the 80s, and he did not say "who's that?" but replied meaningfully to the point.
Feynman seemed to show good understanding of Popperian concepts.
Yeah, he had an aversion to philosophy generally. But he read Popper, or at least he was familiar with his philosophy.
I enjoy Feynman, too, btw. It’s not a personal thing. He made a couple of simple but important mistakes. Inductivist mistakes, no less.
Though I should add that I don't know what he means by "definite" theory. One that isn't vague? Or perhaps one that doesn't require endless computation, as he explains later? Not sure.
Alas, even he made mistakes in his understanding of the creation of knowledge: "[W]e always try to guess the most likely explanation." cringe
Or, less obviously, a bit later: "You can always prove any definite theory wrong."
Sorry? I agree with the first sentence, but lost you after that.
I don’t think so, some wrong answers may still contain truth. Eg Christianity’s commandment not to kill.
From history we know all too well that some were enslaved even though they were much more obviously people than AGIs.
Not to mince words here, but it's the universality of people that implies that everybody is qualitatively equal.
All our laws should apply, but I fear they won't because most won't realize that AGIs are people.
Yes; horribly, that's what companies like @OpenAI and @DeepMindAI seem to be after.
No such thing as "advanced" btw, they are all people just the same.
That's just a map of US airline routes over South America, no?
@ReachChristofer @ToKTeacher @FallingIntoFilm @RatCritical @Malcolm_Ocean @DavidDeutschOxf @reasonisfun @Crit_Rat
Thank you, good point... not sure. Interestingly, whenever subconscious problem solving is successful, the solution does suddenly jump into consciousness (eg shower thoughts etc). So maybe it's something about the correction part of error correction...
@ToKTeacher @FallingIntoFilm @RatCritical @Malcolm_Ocean @DavidDeutschOxf @reasonisfun @Crit_Rat
... until I ride the bike completely subconsciously. Perhaps consciousness is either strongly correlated with error correction, or it may even be error correction.
@ToKTeacher @FallingIntoFilm @RatCritical @Malcolm_Ocean @DavidDeutschOxf @reasonisfun @Crit_Rat
Lately I have wondered if one is aware/conscious of wherever one is trying to detect errors. I learn to ride a bike: very conscious of it, I make mistakes all the time. Then gradually as I iron out the mistakes I grow less conscious of it...
@jamessseattle @ToKTeacher @Crit_Rat
(error correction being the primary ingredient of intelligence)
@jamessseattle @ToKTeacher @Crit_Rat
It can't be analog btw because error correction can only happen in digital systems.
And again, a single Turing machine can simulate multiple Turing machines, so parallelism is incidental at most.
@jamessseattle @ToKTeacher @RatCritical @Malcolm_Ocean @DavidDeutschOxf @reasonisfun @Crit_Rat
In order:
Could be but unimportant - single Turing machine can simulate multiple Turing machines.
Processor and memory.
Doesn't matter/is incidental (if even true).
Yes (if you mean spoken language).
@jamessseattle @ToKTeacher @RatCritical @Malcolm_Ocean @DavidDeutschOxf @reasonisfun @Crit_Rat
No such thing as virtual computer (if by "virtual" you mean "abstract"). Computers need to be physically built.
@ToKTeacher @jamessseattle @RatCritical @Malcolm_Ocean @DavidDeutschOxf @reasonisfun @Crit_Rat
Can you explain why its being analog or digital has any bearing on this?
@caerwy @MatjazLeonardis @DavidDeutschOxf @RatCritical @Malcolm_Ocean @ToKTeacher @reasonisfun @Crit_Rat
I don’t think the presence of understanding (ie knowledge) is indicative of consciousness.
It’s trivial to write a function that represents understanding of a prime number.
I guess that consciousness is related to error correction.
@zarzuelazen
This is the kind of vacuous nonsense that has earned philosophy its bad, navel-gazey reputation.
I agree that we start with conjecture and can then test against brain activity. Thank you; you have helped me realize something important about neuroscience.
Then how can we hope to reconstruct the software that caused these patterns? I think there are infinitely many pieces of software that would result in the same pattern.
@recursus @RatCritical @Malcolm_Ocean @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
Yes, we're in agreement here; though I had already agreed that architecture influences speed. But let me ask you this: can two different algorithms, when run, result in the exact same movement in hardware?
@recursus @RatCritical @Malcolm_Ocean @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
Do these performance characteristics not lie in the algorithm itself? Would X not also take 10 years to run on a desktop computer? (Assuming that computer would have the same memory and processing power as the brain you're comparing it to.)
@recursus @RatCritical @Malcolm_Ocean @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
Or to clarify (something I should have clarified before): hardware architecture can and does influence speed and processing power. But it doesn't qualitatively change anything about which algorithms the universal system can run.
@recursus @RatCritical @Malcolm_Ocean @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
Yup, agreed.
@recursus @RatCritical @Malcolm_Ocean @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
"not all algorithms that solve a given computational problem are equally efficient or robust"
Indeed, because this concerns the architecture of software.
@recursus @RatCritical @Malcolm_Ocean @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
"Not all architectures are equally good for running a given algorithm"
A universal computer, no matter its architecture, can run any computable algorithm (within its memory constraints).
@Malcolm_Ocean @RatCritical @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
When a computer breaks, shit can get weird, too: the fan keeps running, or the housing gets really hot, or it randomly flashes bright colors, some keys work while others do not, it keeps beeping for no apparent reason, it runs all programs fine except the calculator app... etc.
@Malcolm_Ocean @RatCritical @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
It does not. Evolution only optimizes the ability of the gene to spread through the population.
@Malcolm_Ocean @RatCritical @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
Evolution does not optimize for efficient ways to organize any alleged modules.
@RatCritical @Malcolm_Ocean @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
If we knew how to program consciousness and ran it on a computer made of chewing gum and vacuum tubes, those interested would start studying the properties of chewing gum in order to understand consciousness.
@RatCritical @Malcolm_Ocean @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
Somewhere in the brain there is memory, and somewhere there is a processor. Like in all computers. So what?
In order to understand brain functionality, one needs to understand the software that's running on the brain.
@RatCritical @Malcolm_Ocean @DavidDeutschOxf @ToKTeacher @reasonisfun @Crit_Rat
I haven't read it, but the problem with learning about the brain's functionality from its architecture (hemispheres, regions, parts, etc) is this: the brain is a universal computer. Since it's universal, its architecture does not matter.
I’ve only read one of his stories, and remember it being alright. But I love the video games by @frictionalgames which are Lovecraftian.