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
@connectedregio1 @ks445599 @RebelScience @bnielson01
I want to understand how the mind works and then recreate it as a computer program.
@connectedregio1 @RebelScience @ks445599 @bnielson01
I think trying to emulate the hardware is a waste of time. We’re trying to simulate the software, and that software can run on computers that are not the brain. So what could the brain possibly tell us? It’s like studying computer hardware to understand web browsers.
@RebelScience @ks445599 @bnielson01 @connectedregio1
Intelligence is a universal ability. Therefore, AGI and "human-level" AGI are the same. Both are people.
@RebelScience @ks445599 @bnielson01 @connectedregio1
We don't need any hardware at the moment. We first need an explanation of how intelligence works. And then we can draw conclusions about its performance characteristics.
Worrying about scaling hardware is premature at this point.
@RebelScience @bnielson01 @ks445599 @connectedregio1
Are you suggesting these things are the same thing? They aren't.
Yes, it's possible to separate them. Software is substrate independent. I wouldn't really bother with the brain at all.
Excellent. I comment on "thousand brains theory" and HTM in the book, check out chapter 7, section "Neuroscience."
And some program in your brain instructed you to say "an apple fell from the tree" just then, no? On the appropriate level of emergence?
It seems to me you're still assuming the program would simulate some reductive state of that process. It doesn't need to. It can reflect the same level of emergence as "an apple fell from the tree." We know this from computational universality.
It's hard to answer this many questions on Twitter. I recommend moving this to critapp if you want to go deeper.
@RebelScience @bnielson01 @ks445599 @connectedregio1
No. Computing is about instantiating abstractions and their relationships through physical objects and their motion.
@RebelScience @ks445599 @bnielson01 @connectedregio1
It follows from computational universality that no such reinvention is needed. Our computers can already simulate intelligence.
@RebelScience @bnielson01 @ks445599 @connectedregio1
Who cares? You won't win a factual argument by trying to establish status or credibility, or by impressing others with your achievements.
That being said, could we use Turing machines instead of Lambda Calculus? Of course, viz. universality of computation. But it wouldn't be as illustrative.
The universality of computation in itself doesn't advocate for or against any particular level of emergence. It works on any level of emergence, and we can choose the one we find informative.
Lambda calculus is not a description of the mind, and I don't use it as such.
But, as to why I like Lambda Calculus to bridge the illusory gap between philosophy and software engineering: because each part of a function maps exactly onto how explanations work.
Either would work; in the former case, you do it yourself through thought, in the latter, someone else through writing code.
@RebelScience @ks445599 @connectedregio1 @bnielson01
Once again you ignored my q :) Can you slow down pls?
Here are examples that would change my mind. A good formulation of a principle of computation showing (sensory) inputs are necessary for computation. Or, more generally, a refutation of the universality of computation.
@RebelScience @ks445599 @connectedregio1 @bnielson01
You keep saying that, despite evidence and good explanations that this isn't the case.
What could someone possibly say that would change your mind about this? In a rational discussion, one should be prepared to answer this question. (It's not a rhetorical q, I'd like to know.)
@RebelScience @connectedregio1 @bnielson01 @ks445599
So in other words, you want to keep ignoring my questions, and you want to insist that computation is impossible without inputs, even though I showed you a routine example above that performs just that?
If you want to make progress here, it's high time to change your mind.
@RebelScience @connectedregio1 @bnielson01 @ks445599
Can you answer my previous question?
And, if you are right, can you please explain how the function I posted above performs the requisite computation even though it doesn't have any inputs? Notice how the brackets behind "foo" are empty; that's where inputs would go.
Agreed. And software engineering can help improve both, including mental ailments that are the result of bugs in the "user space."
What difference does it make whether and how people discuss these things? :)
In any case, I recommend chapters 3 and 6. Especially in the latter one I address criticisms similar to the ones you propose.
@RebelScience @connectedregio1 @bnielson01 @ks445599
In other words, you're saying that nobody could possibly come up with an argument for why you should change your mind about this?
@connectedregio1 @RebelScience @bnielson01 @ks445599
(and the vast majority of our computers are; our laptops, smart phones, desktop computers, etc.)
@connectedregio1 @RebelScience @bnielson01 @ks445599
Their brains are. But yes: whatever happens in a human brain - creativity, among other things - must be replicable on a computer other than the brain, if that computer is universal.
@RebelScience @connectedregio1 @bnielson01 @ks445599
"Data" means "givens." That function I sent you does not take any inputs: it's not "given" anything, and especially not through senses. And yet it performs computation, which is something you claimed is impossible.
@RebelScience @bnielson01 @ks445599 @connectedregio1
There are whole programming languages that are built around the idea that code is data. They are called "Lisps."
What's the basic machinery of minds?
No. Software engineering is not an appeal to reductionism. It happens on the same level of emergence on which ideas live.
@RebelScience @connectedregio1 @bnielson01 @ks445599
As to your point "there can be no computation without data." Here's a simple program that takes no data:
(defn foo [] (* 2 2))
It returns 4. Are you saying this isn't computation?
@RebelScience @connectedregio1 @bnielson01 @ks445599
Organisms are born with data that's genetically given. And an organism would retain all of that genetic data even if born with a malfunction that cut off its brain from all sense data.
@connectedregio1 @RebelScience @bnielson01 @ks445599
A computer is universal when it can compute anything any other compute can compute. A universal computer can compute any computable function.
@RebelScience @bnielson01 @ks445599 @connectedregio1
No. Code is data and remains data even if cut off completely from the outside world, with not ability to ingest additional data.
I'm guessing this mistake rests on a misunderstanding of computational universality.
By the same logic, someone who is, say, schizophrenic, just needs to be talked out of it?
The comparison doesn't hold because neurosurgery and software engineering happen on different levels of emergence.
It's still possible it would need to update some parameters at first, which may take some time, and so it might initially stumble about a bit. Progress might look like "learning" - in which case, such a result may not tell us much.
2/2
Off-the-cuff guess: it would do better than a human initially, because presumably its knowledge of what to do with visual impressions is given genetically and remains present even when blind. Once the "veil is lifted," it just needs to invoke the knowledge.
1/
I found the video of the dog playing Jenga, if anyone wants to see it:
It was fun being on Christofer's show again. twitter.com/ReachChristofe…
@bnielson01 @ks445599 @connectedregio1
I call them "functions," but all replicators (including memes and genes) are functions, so I have been looking for a new term as well! If you think of one, let me know...
@RebelScience @bnielson01 @ks445599 @connectedregio1
Well, we also know from the universality of computation that we could build general intelligence on computers that don’t have any sensory inputs.
Either refute the universality of computation or stop invoking special-purpose brain regions and sensory inputs.
@RebelScience @bnielson01 @ks445599 @connectedregio1
No because we know from the universality of computation that we could build general intelligence on computers that don’t have visual cortices.
@RebelScience @ks445599 @connectedregio1 @bnielson01
Someone should write about why thinking about brain regions is pointless in this regard... Oh wait, I already did: amazon.com/dp/1734696109/…
@bnielson01 @connectedregio1 @ks445599
I suggest you move this to critapp if you want to have a more detailed discussion. It's near impossible to have long-form discussions on Twitter.
@bnielson01 @connectedregio1 @ks445599
We should take a step back. The underlying issue here is that you ignore my questions and that you are still an inductivist without realizing it. After ignoring a question you usually follow up with long, somewhat unrelated arguments.
@bnielson01 @connectedregio1 @ks445599
Again, why would there be nothing to learn?
@bnielson01 @connectedregio1 @ks445599
People don’t learn from their environment.
@bnielson01 @connectedregio1 @ks445599
In fact, one doesn't need access to any environment for learning to happen, let alone a random one. A brain in a vat can still create explanations.
Thinking that some environment is needed, or that it need contain regularities (non-randomness), is just a cousin of inductivism.
@bnielson01 @connectedregio1 @ks445599
Why would there be nothing to learn?
People may well guess the explanation "things happen randomly," and they'd be correct. By creating this explanation, they would have learned something.
And by creating mistaken explanations and then improving upon them, they'd also learn.
@bnielson01 @connectedregio1 @ks445599
Yeah, which makes me think the brain must be the one non-random thing in this scenario.
But no, I don’t think it would make every explanation coincidence. That’d be judging explanation by likelihood, no?
Not all animals, but many conceivably have both. It’s harder not to evolve it. And our ancestors, which were animals themselves, must have had both for us to evolve intelligence.
Though I agree with the conclusion, many animals’ brains may be universal computers.
Universal computation is a necessary condition for intelligence, but not sufficient. What’s needed in addition is explanatory universality.
Why wouldn’t people be able to develop explanations in such an environment?
They would wonder about their environment, and try to explain it, no? And maybe they would conjecture that it’s random.
Assuming 4 through 7 are true, aren’t they part of 3?
And isn’t taking vitamin D better than direct sun exposure?
I agree.
Effective psychotherapy is a branch of software engineering. Present-day approaches happen on the level of the brain, not the mind, and, therefore, are ungainly at best, counterproductive at worst.
Developing effective psychotherapy is a big part of AGI studies.
My book "A Window on Intelligence" is now available worldwide, anywhere you can buy books.
Amazon: amazon.com/Window-Intelli…
Apple Books: books.apple.com/us/book/a-wind…
Barnes & Noble: barnesandnoble.com/w/a-window-on-…
Learn more: windowonintelligence.com
I have been looking forward to this day for almost a year.
Today, a SPECIAL episode for my listeners. Introducing: A Window on Intelligence - The Philosophy of People, Software, and Evolution - and Its Implications.
Be the first to listen to the excerpt: soundcloud.com/dchacke/13-int… https://t.co/eH6BbVsmHg
Since evolution is a gradual process, I guess there was never a single genetically unique organism that wasn't the last of its species (that even includes humans).
Entire species can be ~unique if their last common ancestor is sufficiently far back up the phylogenetic tree.
@dela3499 @KittJohnson_
Indeed. Likewise, the gaps in human thought are filled with viable ideas. Not necessarily viable vis-a-vis a problem, but vis-a-vis spreading through the population of a mind's ideas.
Otherwise, we need to explain evolution as anything other than gradual.
@dela3499 @KittJohnson_
Yes. And are there not dramatic genetic gaps between, say, a sea horse and an elephant?
@dela3499 @KittJohnson_
Perhaps :) Or, to make things easier, space shuttles. But maybe even those could be evolved biologically in principle.
Also, the reason human thought can escape parochialism is not that it can jump gaps - it's still gradual - it's that a human with bad ideas doesn't die.
@dela3499 @KittJohnson_
I guess that Lambda Calculus is evolvable, and that evolution stumbled upon it pretty soon, because it is easier to evolve than any non-universal counterparts.
I guess that is true of other universal models of computation, too.
Conversely, what isn't evolvable in principle?
Another example of buggy animal programming (one that many will, presumably, readily interpret as intelligence, despite being evidence of the opposite). twitter.com/jimcaris/statu…
@mizroba @DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
Through a population of memes, yes.
Count me as one of those silly people who doesn't think animals can learn :) At least not in the sense of creating knowledge - animals that copy memes do so through imitation and don't seem to create knowledge in the process.
@mizroba @DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
Not a population, but a kitten not grooming itself until exposed to a cat who does is an example where memes are a better explanation than genes. For, if the kitten had contained the knowledge of how to groom genetically, why didn't it groom? Maybe that needed to be "activated"?
@mizroba @DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
Ah, I'm guessing that's what you meant by "universal" animal behavior: that which thousands of unconnected populations share.
For those, I agree that genes are the better explanations. Nonetheless, some animals do have memes.
Do you agree?
@mizroba @DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
It can. Memetic evolution may independently converge onto similar solutions, as evolutionary algorithms often do.
(It is for this same reason that a rabbit fossil in Cambrian rock would not refute the biological theory of evolution, but it applies equally to meme evolution.)
@mizroba @DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
"Memetic" means "of or related to memes." I don't know what you mean by universal animal behaviors, but yes, some animals have memes.
@DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
Top-down causal chains are real, and they happen from meme to DNA, but reductionism denies that and says that causation can only travel bottom-up. So in a reductionist framework, it seems to make sense that genes have all the control. Which leads to many new problems :)
@DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
So then genes presumably have enough control to build in a fail-safe that controls human behavior to the degree that it isn't detrimental to the genes? And so then humans aren't universal explainers after all? No.
And DNA molecules are made of atoms, so do atoms control DNA?
@DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
Ah, our old friend reductionism. :) Genes build the neuronal architectures that store memes, sure, but that doesn't seem to have much impact on the kinds of memes people can have: how do memes such as fasting, homosexuality, etc spread despite being the gene's worst nightmare?
@DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
And so we may consider the dysfunction or disappearance of the DNA molecules that used to encode those genes to be part of that meme’s phenotype.
2/2
@DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
I think so. I once heard that cats’ grooming behavior is memetic not genetic. Suppose it used to be genetic. After the meme of grooming spread through the population of cats, mutations of the corresponding genes occurred and now those genes are either dysfunctional or gone...
1/
@DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
Yes, I'd consider those DNA molecules part of the phenotype of the meme of growing meat in a lab.
Not the genes themselves, though - they are abstractions, and phenotypes are about effects on the physical world.
@DKedmey @BretWeinstein @RichardDawkins @ToKTeacher @DavidDeutschOxf
Some memes are encoded in genes, e.g. the human meme of pointing in some dogs' genes (credit to David for telling me this). In such cases, the DNA molecules of such genes could be considered an extended phenotype of those memes, yes.
Is that the sort of thing you had in mind?
@ReachChristofer @DavidDeutschOxf
I’m guessing that’s how one could demote a person to an animal. (Horribly immoral to do so!)
If we were to delete all ideas from a mind, I’m not sure it’d be a mind anymore, but it may be more of a semantic issue at that point.
2/2
@ReachChristofer @DavidDeutschOxf
If we could somehow delete those ideas from a mind that replicate within it, then evolution in that mind would stop and it wouldn’t be creative anymore. It may still contain non-replicating ideas, though, and therefore be a non-creative space for ideas.
1/
@Giovanni_Lido @noa_lange @DKedmey @DavidDeutschOxf @veritasium @Crit_Rat @LRNR @JohnHMcWhorter
I can't decide if it's better or worse to only have one term for it. :) The German 𝑤𝑎ℎ𝑟𝑠𝑐ℎ𝑒𝑖𝑛𝑙𝑖𝑐ℎ is an interesting case: as an adjective, it means what you said. As an adverb, it seems to mean "probably," e.g. "wahrscheinlich richtig" means "probably true."
@DKedmey @DavidDeutschOxf @veritasium @Crit_Rat @LRNR @Giovanni_Lido @JohnHMcWhorter
(using "likely" as a synonym of "probable" here)
@DKedmey @DavidDeutschOxf @veritasium @Crit_Rat @LRNR @Giovanni_Lido @JohnHMcWhorter
I don't know Greek but the confusion seems to date back to misinterpretations of Xenophanes' usage of the word "eoikota."
3/3
@DKedmey @DavidDeutschOxf @veritasium @Crit_Rat @LRNR @Giovanni_Lido @JohnHMcWhorter
The terms are opposites because the more like the truth a theory is, the more non-trivial, complex, and bold it is, and therefore less likely to be true.
2/
@DKedmey @DavidDeutschOxf @veritasium @Crit_Rat @LRNR @Giovanni_Lido @JohnHMcWhorter
Epistemologically speaking, these sound off.
According to Popper's C&R, the confusion between "likely" and "like the truth" dates back to misinterpretations of Xenophanes in ancient Greece who used the latter meaning.
1/
@tjaulow
I haven't, but I guess behaviorism puts a cap on how much he can contribute to AGI research.
@PopperPlay
Evolutionary algorithms (genetic programming). We don't yet know how to build AGI with them, but they seem to be the only promising path.
@PopperPlay
Glad you like it. Reach as in could potentially help build AGI, or as in learn about functional programming generally?
@joe_shipman @DavidDeutschOxf @ESYudkowsky
I recorded an impromptu episode of my podcast to go into more detail about this:
I recorded an impromptu episode about the question of "value alignment" in artificial intelligence research:
@RealtimeAI @ReachChristofer @dela3499 @ToKTeacher @Soph8B
Not really, no. Here's a hard-to-vary explanation of explanation: youtube.com/watch?v=DHR6ro…
The mistake Popper identified as the who-should-rule-question permeates inquiries into AGI and is behind the pseudo problem of “value alignment.” An example of how critical philosophical knowledge is in this field and how urgently we need more of it. twitter.com/joe_shipman/st…
It’s undesirable because when taken seriously it leads to enslavement and murder of AGI.
2/2
One can’t know, nor does one need to. The very question seeks an authoritative answer (see the only superficially unrelated economist.com/democracy-in-a…; disable JavaScript to circumvent the paywall). And mistakes are what enables progress in the first place.
1/