Hacker News Reader

Comment by jcims | Hacker News Reader

> sift through material from around the body in order to scan for infection. Without lymphatics, fluid would build up in body tissues, and there would be no way to alert the adaptive immune system to invading pathogens

sift/scan/alert/invade

I always find it interesting how much agency we assign these systems when we describe them. Listen to some of the descriptions of our genetic machinery, you’ll hear the same thing. I understand much of it is to help the audience understand, but even when you dive deep into a topic it doesn’t seem to go away entirely.

View on HN

I find this semantic critique interesting but I have to disagree with you. Underlying the critique is a hidden arrogance.

Agency language is strongly tied to our ability to understand the mechanism by which phenomena occurred. For example, take 4 different chess bots:

1. The bot moves in a 100% predictable way to you (say it checks each piece in order to find a possible move, the first it finds that's legal it moves)

2. The bot moves randomly

3. The bot uses minimax + library (traditional chess engine)

4. The bot is running via reinforcement learning

When we discuss the moves the bot made, in which of the above scenarios would we say it 'chose' to move in a particular way? These different examples call out different forms of 'agency'. In cases 3 and 4 (maybe even 2?) most people - including the creator of the bot - would say the bot chose to move pawn to c4.

As effect gets sufficiently far removed from cause it's natural to ascribe agency to the mechanism. In a very real sense in case 4 the creator designed the bot but has no idea why it moves in certain ways - they can analyze the weights of course - but why did it 'choose' those weights over others during training? If we knew that we wouldn't need the costly training in the first place.

So back to your distaste of agency language when describing a phenomena. We're talking about _incredibly_ complex machinery. There is no guarantee that these mechanisms are a 1. or even a 3. type bot. They could be a 5. or 6. (whatever that even means). Thus even if we knew _everything_ about the last order mechanism, we would still ascribe agency to the effect.

>So back to your distaste of agency language when describing a phenomena.

Sorry rereading my note I can see why this seemed like a critique. I really mean 'interesting' in its standard denotation.

We're apes. Our basic assumption is that everything is also an ape. Including electrons, and lymph.

It's very hard for us to find any other kind of explanation intuitive.

We routinely use terms like "dependency injection" for programming, "drop the beat" for music, "oscillating" for a set of political views and how they develop throuh time, or even "transcribing" for DNA-related activities in the cell, if you want a biological one. None of those have to do with apes or their everyday lives. Or try explaining "political correctness" to an ape...

We are perfectly fine abstracting high-order things that have nothing to do with apes, and using those terms and abstractions IF they are useful. We are also perfectly fine to not bother with it if existing terms offer the best possible utility.

So no, our basic assumption is not that everything is an ape.

I'm not sure where this fascination with undermining the great intellectual feats human brains are capable of is coming from?

> our basic assumption is not that everything is an ape. > great intellectual feats

I think it is, exactly, a feat of the imagination to arrive an an explanation which does not require this premise.

But it is not our default presumption. Hence 100,000 years of superstition, and 500 years of rare scientific moments.

Concepts seeming abstract, is in some sense, a bit of an illusion. Concepts are, in my view, always psychomotor. They succeed in "abstracting" from our bodies, sure, by metaphor. But the genealogy of that metaphor is the psychomotor system of an ape.

Consider your words: injection, drop, beat, oscillate... these are about how our bodies behave in (with, towards) our environment; and become "abstracted".

I think there is something quite fundamentally apeish when we think of a song dropping a beat, in the sense that our relatedness to the song is actually facilitated by an apeish psychomotor state of "dropping something".

Yes, we can reduce our engagement with some phenomenon to a purely cognitive model (pieces and their relations); but that is very hard, and typically is only useful is narrowly solving known problems.

Innovation (, creativity, etc.), including the scientific, relies quite heavily on our being able to impart to alien things some intuitive apishness which we then, at best, abstract away.

Many of course cannot perform this final abstraction. Which is why we speak of "The Economy" and other such apelike characters of the imagination.

If you turn your argument around, almost every example can be used to support a different hypothesis:

our terms and thinking is based on abstraction, a certain feat of processing information. Other earlier Great Apes and even other animals are simply less able to implement this abstraction, and because of this they only use the abstraction for computation of their simple lives like motor skills or "psychomotor states". Since we have more brainpower, we are able to implement a higher percentage of abstractability.

Thus, the core aspect is the program of "abstractability" (or it would be more common to use the word consciousness or awareness). We implement it to a higher degree than other apes, but since we both implement parts of it, of course we are going to share some aspects of it in our lives.

> Concepts are, in my view, always psychomotor

Even this seems much more accurate to view the other way around: both psychomotor skills and concepts require computation and abstractability. Concepts are always based on abstraction. But psychomotor skills are also based on abstraction.

And again, what does "political correctness" have to do with a psychomotor engine of an ape that can jump between trees, hold a balance and throw things at a distance? You would need to use some serious imagination to convince people that the most occam's razored hypothesis of that concept is based primarily(!) on mechanical skills?

But I guess hey, we won't know for sure until we make AGI so who knows ;)

In a sense both views are correct. Our psychomotor states instantiate the abstraction which are drawn from them (or else, why abstract?).

eg., there is a level of abstraction at which an ape "alerting" another ape is the same as an chemical "alerting" a cell

My point is that these things are only equivalent to these abstractions in a very narrow kind of activity (cognition).

A concept in the brain-body of an ape is a means for coordinating itself with respect to its environment: coordinating its eye, its hand, and latterly, its mind (, emotions, thought patterns, etc.). (NB. political correctness is here just a regulation on how we socially coordinate).

That's just what we use concepts for.

They also play a purely cognitive role in the sense that "circle" is both a guide to the motion of my hand in drawing a circle, and also a "Form" to which cognition can be purely and abstractly directed (ie., a shape with various properties...).

But this cognitive role is derivative of their non-cognitive origins in, quite literally, structuring the meat of our bodies. It is this structuring process which is how we "conceptualise" the world.

A dog is replete with many concepts; it is the dumb cognitive illusion of apes lacking self-awareness to think their linguistic cognition is where "concepts" lie. This is a cognitive shadow of how every animal navigates its environment.

Your "AGI" remark indicates to me that you may take this cognitive view. I think this is the consequence of a mathematical discipline (computer science) being "let into the room" -- it should be thrown back out.

Mathematics hasn't much to say about intelligence. The dumb walnut we call a frontal lobe possesses no models of how it actually works, and only the shadow of concepts it calls "abstractions". It is a delusional Cartesian place with little to offer self-understanding.

Everything important for the development of intelligence, etc. will come from the non-cognitive (, bodily) aspects of mental function which will require building animals. Cognitive ai systems cannot conceptualise, they only automate the walnut.

You might have some interesting ideas but I am having a hard time following them to see a point you are trying to make.

Also you constant usage of the word "dumb" or "delusional" confuses, it looks like you apply some personal judgements onto parts of the discussion where it has no reasonable place to be.

So far it seems that you are trying to make the point "This is a cognitive shadow of how every animal navigates its environment.". This interpretation seems to me like an attempt by you to fit the data into this non-fitting notion that everything is based on psychomotor processing, because your mind is in love with this idea. And it's a beautiful idea, I agree. But I don't think it's the most truthful one in this space.

> (NB. political correctness is here just a regulation on how we socially coordinate).

The question was, how specifically is it rooted in motor cortex or psychomotor functions? To explain that you will likely have to widen the definition of "psychomotor" to such an extent that it would lose all its useful meaning. Which is exactly one of the signs, for me personally, that your theory is only making very limited sense with regards to everything we know about brains and minds and behavior of consciousness. That it's a beautiful theory that unfortunately too thin to fit the real world in it.

You touched on an interesting thing regarding AGI. In the purely practical sense, I see that a lot of people have taken this apparent rule(or a "universal truth") of making an AGI that says "a mind requires a body to even function or be able to be aware"-thing. People think it means actual bodies. In reality they can very likely be implemented as virtual bodies or just address spaces that can interact with their environment in an adressable way. Thus, a "body" is just a part of the AGI, a computation structured in a certain way to make sure the rest of it works properly. Like a generator in a GAN network.

In very simple terms, the machines' bodies can be just avatars in some video game, it would be enough to satisfy the requirement of the aforementioned rule, if game connects all the inputs in the correct way.

With "dumb" i wish to debase and defuse the pretentions of the intellect.

And I dont mean your intellect. I mean the thing we call "reasoning" towards which we can draw our awareness.

This is in some profound way, quite dumb. It seems smart for all sorts of reasons: in particular, it is furnished with concepts that have taken millions of years of cultural evolution to develop.

An ape, such as us, absent this cultural context would seem precisely as dumb as it really is: very dumb indeed.

We misattribute our reasoning with these concepts for their inherent intelligence. This intelligence, which we might call conceptualization, is not a congitive process. The real intelligence is animal intelligence, it's the bedrock of our cognition which is always outside of it; and cannot be observed from within.

> Thus, a "body" is just a part of the AGI,

Not at all. A body is just the place the world changes. It's the thing in direct causal contact with the world; whose effect is a kind of painful restructuring. eg., the bike rider learns to ride a bike not by "evolving his cognition", but by the structure of his muscles changing.

(Indeed, by his eye swirling differently in its socket).

It is this restructuring which is an essential aspect of conceptualization (whose effect we see in consciousness).

I strongly suspect that this kind of restructuring is necessarily organic: it is growing differently given the impact of the world. I am open to other systems which can evolve their physical structure in this way, certainly, i do not think sand qualifies.

re Political Correctness

Correctness = straightness, fit Fit = environment-body "matching"

Political = other bodies wanting from me, my wanting from them

political correctness = criteria for the fit of my body with other bodies as we want from each other

In being a criteria it is a "concept about a concept" and therefore its heavily on the psych-side of psychomotor, ie., it is a kind of regulating mechanism on reasoning (specifically, reasoning about my fit in an environment of people).

> This intelligence, which we might call conceptualization, is not a cognitive process. The real intelligence is animal intelligence

I can only partially see what you mean exactly by "cognitive" or "animal" (with regards to intelligence), it doesn't seem like you use the words in their precisely common meanings.

I also never claimed that intelligence is in its base cognitive.

The original claim was that all of the intelligence is based in "ape-ish" motor skills, that it's basically their motor cortex that is just a little more advanced that gives rise to all the behaviors that we call intelligent.

This idea would basically mean that a task such as writing and publicizing a great book such as Shakespeare's Hamlet (including accumulating all the life experience required to produce it, and make it being so good) - is "basically a motor skills task".

This just doesn't make a lot of sense to me.

While it does of course require motor skills like taking a pen and putting the words to the paper, it requires a lot more other complex skills as well that have so little to do with motor intelligence that I don't see any honest reason to put them in the same category.

Thus, my comment that your theory requires expanding the concept of "psychomotor" to such extent that the word loses all of its useful meaning.

> political correctness = criteria for the fit of my body with other bodies as we want from each other

It's just not a good explanation, it loses 90% of all of the intricacies. Try using it to explain it to someone who doesn't know what political correctness is- it doesn't do the job.

> cognitive

I just mean any system of inference over propositions, or anything which looks like it .

We are told our frontal lobe is the most unique aspect of our brains: not true! Our motor cortex is.

We are the only animal which learns almost all skills after-birth, via our significant and adaptable motor cortex.

> Try using it to explain it to someone

Yes, concepts need to be converted/reduced to propositional (and then linguistic) forms in order to be cognized and then to be communicated.

You can't really acquire a novel basic concept via communication, but you can acquire metaconcepts (like criteria for things) which are just combinations of basic concepts you have already learnt (ie., motor concepts).

I'm not saying that there is no propositional content to concepts, ie., that they fail in being instances of some abstract Form.

I am saying, rather, what they are; and how we acquire them.

I have significant doubts that the world can be conceptualised by mere cognition, ie., that the contents of an inferential process can be arrived at via an inferential process.

I think this is an illusion: and it is apparent with machines. There is no mechanism for their understanding what they are processing.

We aren't aware, i'd say, at all, of how we actually conceptualise the world. All we're aware of, as far as our intelligence goes, is inference(/cognition). (Though many greats have attributed mysterious immediate insights to the divine: rather it is just their actual intelligence).

I think that sets us up for a highly misleading paradigm of intelligence (cf. Cartesian dualism, etc. )

Nothing to add apart from the feeling that you have shown complete mastery of how to explain something.

Is there any problem with it? That's just how the language works. In many of the contexts and on many of the logical levels the body does indeed have agency: the linguistic expression embedded in agency terms and all related expressions are just the most accurate and compressed and useful descriptions of phenomena we are studying. They have all the useful connotations and sub-meanings that help the process of expanding knowledge forward. All I see in those terms is best possible ways to explain the actual found relationships between parts of the immune system, instead of some description of meaningless facts or chemical reactions.

If you insist on being on some imaginary laplace daemon lowest level of interaction with the world, maybe theoretically you could escape the agency component (but even then you have problems with questions like "what makes all particles interact with each other at all"), but nobody lives on that level anyway. Just because in some distant theory it is possible to reduce everything to interaction of some miriad of particles with exactly known states and rules, it doesn't mean that you (or other human or society) can effictively and practically compute anything of value on that level of abstraction. Nobody interacts with life on that level. No useful discussions are had on that level in most of the scientific pursuits.

This is a matter of using the language effectively, not a matter of superstition. I don't see scientists cargo culting the immune system or organs of the body.

Think about it like this: if I'm programming in Python, I'm going to think about variables, scopes and closures, not about high and low voltage levels or number of silicon atoms in my processor. A closure is not an illusion just because there is no transistor that specifically makes it work, a closure is still the most useful term on it's own level of abstraction.

(Recommended reading: Gödel, Escher, Bach: an Eternal Golden Braid, also known as GEB, a 1979 book by Douglas Hofstadter)

>Is there any problem with it?

No problem at all. Re-reading it there was much more snark than I intended (which was 0)

When talking about the immune system there's tons of agency.

There is an essay[1] from last month, which argues for assigning agency to biological systems and not just for illustration for the general public but for research.

For me, a layman, it had some convincing arguments, at least in abstract.

[1] - https://aeon.co/essays/how-to-understand-cells-tissues-and-o...

This reminds me of a point my physiology instructor worked hard to hammer home in the senior year of my bio degree: do not assign agency to unconscious processes. I forget the rationale for this but find myself reflecting on it often.

Reply on news.ycombinator.com