Hacker News Reader

Comment by janalsncm | Hacker News Reader

What amazes me about spiders is their complex unlearned behavior. No trapdoor spider learns how to build a trap. No orb weaver learns how to spin a web. Trap/web building behavior is entirely programmed in their DNA.

I think about this a lot in the context of conversations about intelligence. If spiders can have complex behaviors hard-coded, humans certainly do too. (In other words, the Tabula Rasa theory is wrong.) The ability to learn language and emotion are certainly two examples. We are pretty good at certain things (learning language, picking up social cues) and relatively bad at others (calculating an 18% tip).

So if you’re going to measure “intelligence” the first thing you’ll need to do is choose what to measure. What questions do you ask? You might be inclined to pick things we humans think are important. But then that’s not an objective universal measurement, at best it’s yardstick for human cognitive abilities.

View on HN

Story time.

I'm from Western Australia and spent a lot of time in its deserts in my youth.

I have a very distinct memory of walking out among the spinifex as a child, needing to pee. Its hot, my bare feet are burning, the ground is unforgiving and I already have spinifex'ed ankles, so naturally I tread as carefully as I can, trying to avoid the sand and the spikes, only walking on the slightly less heated organic mats that surround some of the spinifex, a kind of carpet of twigs and leaves, gathered in pockets by the wind.

The natural urge is rising, and I stop paying attention and get a bit of pace, as relief must be had. I see a particularly appealing area, covered by this organic dust, and I leap over to it, get my body parts aligned, and begin the process of relief.

A few seconds into it, I feel something moving. The entire mat that I was standing on, is in fact, a field of trap doors. Maybe a hundred of them. They begin to open.

I have never levitated as hard and as fast in my life.

So my Tabula Rasa has given me this lesson: don't go into the spinifex to pee. Ever.

Wild. I also grew up in Perth and until just now thought WA didn't have trapdoor spiders. Thanks for the anecdote that taught me something.

In fact they recently found new species ...

https://www.australiangeographic.com.au/news/2018/05/fourtee...

I wonder if the process of building a complex web can be reduced to a simple set of rules, almost like a cellular automata.

The simple rulesets which lead to complex outcomes in nature is amazing to me.

simple nets for sure, but spiders can easily deal with an uncountable number of corner cases, I somehow think it's not reducible to all-too simple rules (in the 'game of life' sense) and more an interaction of several non-trivially complex systems like their various perception systems.

obviously their neural networks are limited in size and therefor complexity has to remain somewhat simple relatively to larger animals).

https://xkcd.com/1135/

Many years ago when I was working IT for a public school district, I was unpacking a large number of identical HP PCs from cardboard boxes. I noticed that a large portion (it must have been half or more) of the boxes had the exact same type of spider in a nearly identical web in the exact same spot in the cardboard box. It was a small cardboard cubby taped onto the inside of the large exterior box, for storing a power cord and other small accessories. It must have been just the right shape and size, and accessible in just the right way, that this type of spider absolutely loved it.

> If spiders can have complex behaviors hard-coded, humans certainly do too.

Depends what you mean by complex... All creatures get born with an idea how to get food. Human babies have never seen breast or anyone sucking breast, yet they can go through the process pretty much immediately. That's a very clever movement / sensation following combination, but it's there from the beginning.

do you have children or worked with babies? they suck anything they can fit in mouth. The reflex of sucking is wired, yes, but not the object

Babies are definitely hardwired to be attracted by the contrast of skin / areola on the female breast. Once they have targeted this contrast, they will do a very typical (also hardwired) "search" for the nipple: they come very close to the breast, open their mouth, do exploratory sucking via fast breathing, and quickly move their head left and right until they have "docked" on the breast. They start doing this minutes after birth (they are also hardwired to search for the breast when placed on the mother right after birth), and stop once their sensory systems are fully developed and they have learned how mom looks, how mom smells, and where mom's breast is (after 1-2 months). On my own child, I even noticed how it once mistook the snout of a teddy bear for a female breast.

The "will suck at anything" instinct is IMHO a fall-back mechanism if for some reason a female breast is not available.

Here's a source for the above claim that I found after looking this up:

>During pregnancy, the nipple has become more pigmented 68 and is easy for the newborn infant to discover (Fig. 1C). We have observed that soon after birth, the areola expands and takes a bulb-like shape (Fig. 1G). The Montgomery glands also become more pronounced (Fig. 1G). The scent of areolar secretions has been linked to behavioural responses, such as head turning 69 and directional crawling in newborn infants 70. This release of the breast odour by the Montgomery glands is known to help the newborn infant find the nipple 17, 27, 69. The newborn infant recognises the scent of the mother's breast from the amniotic fluid 71, touches the breast and transmits the taste of the breast to the mouth (hand-to-breast-mouth movements) 17. This stimulates rooting and crawling movements in the newborn infant to reach the nipple. The connection between the taste of amniotic fluid and the scent of the breast from the Montgomery glands highlights a biological survival mechanism – a pathway of flavour with lifelong consequences. (Fig. 3) (17, 71-80) When knowing of this sensitive odour-dependent mechanism, it might be wise not to interfere with unfamiliar hands.[0]

[0] https://onlinelibrary.wiley.com/doi/10.1111/apa.14754

There's more to it. Have a look at the specific idea of "breast crawl".

I kind of wonder whether certain cultural artifacts might be genetically inspired. My son, when he was young, without any prompting wanted to be an architect (like his great-grandfather whom he had never met, nor even heard anything about before this) and was obsessed with guitars (like me, although at that point, my own guitar had been in the basement for years and he had never seen me play it). This is not to say that these particular behaviors are explicitly encoded in genetics (both guitars and architecture as a discipline are much younger than evolutionary time), but rather that some intrinsic predisposition towards them exists (so, perhaps had we lived in fifteenth-century Persia, we would be oud-obsessed, to choose one posseible alternate reality).

Anecdotally, I had a grandfather I never met who had built his own house on a nice plot of land out in the sticks, and that's something I'm obsessed with doing in the future. My father also worked in tech, but didn't really pass knowledge to me.

Though, both of these things are almost stereotypical [1]

1: https://www.reddit.com/r/ProgrammerHumor/comments/vnmmuo/whe...

I'd hazard:

- fascination with fire, - love of tools, - teenage interest in weapons.

> If spiders can have complex behaviors hard-coded, humans certainly do too.

That does not follow. It's a logical fallacy. ('Spiders can have eight legs, so humans certainly do too.')

https://en.wikipedia.org/wiki/Polymelia

That's a birth defect, not applicable.

According to scientific evolutionary dogma, all mutations are birth defects.

Shit, all life is a birth defect? Explains a lot.

in the limit the distinction between something that is learned and something that is programmed genetically is somewhat a matter of definitions. if a species genes makes it so they create a complex social web that teaches all its individuals certain skills, is that genetically programmed or not?

free will wouldn't even change that, since it's presumably also given by our genes.

It's only a logical fallacy if it is taken to be the entirety of the argument rather than an abbreviated representation of a longer chain of reasoning. I don't think anyone would claim that sentence is self-evidently true!

Try this one: we observe complex instinctual behaviors in most animals on this planet. Therefore it follows that humans are special and have no complex instinctual behaviors.

We struggle to define intelligence, much less measure it.

When I was at school a million years ago "smart" kids were ones who learned to read fast, did numbers well, and had good memory recall.

Then along came computers, which had perfect memory, did numbers really well. So "intelligent machines?"

Obviously not, so we tweaked our measure of intelligence. Perhaps to include writing well, or learning quickly. Which, um, computers now do.

But we still view the computers as "not intelligent" but I'm not sure what measure we'll use now. Something to do with art or music? Hmm, no, wait, that's no good either....

(PS schools still mostly measure numbers and memory....)

"But we still view the computers as "not intelligent" but I'm not sure what measure we'll use now. Something to do with art or music? Hmm, no, wait, that's no good either...."

We could focus on whether the student has actually understood the topic and is not just able to produce nice sounding words and some quotes about it.

How do you differentiate between a student who actually understands the topic and one who is just able to produce a convincing essay about it?

More generally, how do you measure ‘actual understanding’, if not by judging the student's output?

"if not by judging the student's output"

Where did I say to not do that?

And someone who can produce a convincing essay usually has understood the topic (unless he copy pasted it). And when there are doubts, one can ask that student specific questions.

> Where did I say to not do that?

You asserted that the student should “not just [be] able to produce nice sounding words”. What do you suggest output is if not words (nice sounding or not)?

> And someone who can produce a convincing essay usually has understood the topic (unless he copy pasted it).

Then, what is the difference between a convincing essay (‘nice sounding words’) produced by a human, and similar words produced by an LLM?

After all, even today, students are passing ChatGPT’s work off as their own. Does this then demonstrate understanding by the student or by ChatGPT?

> And when there are doubts, one can ask that student specific questions.

If an LLM can answer specific questions about a subject, does this demonstrate true understanding? If not, why not?

"If an LLM can answer specific questions about a subject, does this demonstrate true understanding? If not, why not?"

If it could, I would say there is understanding. But in every area where I have expertise, I see LLMs failing my questions. Just like a bad student would, who memorized textbooks, but did not connect or process the knowledge in any way.

Schools deem a child smart when that child excels at tasks that the school assigns to them. I've those tasks were simply to read and add. Then those tasks included writing and remembering. What tasks do schools demand of children today?

It seems that intelligence is a term that we apply to people who consistently excel at the mental tasks assigned to them, no matter what those tasks entail.

Kids being smart normally means they are _more_ intelligent than the others. Computers, compared to average humans are dumb as rocks.

Who struggles to do what for which purpose?

Sometimes the word “intelligent” is used by someone who sells IQ tests.

Sometimes the word is used by someone who sells the idea of a computer that can think like a human.

I don’t know if people in general struggle or care about the definition.

I saw a project on GitHub some time ago where someone trained a neural net with only a single, shared weight to play a driving game.

They used a gentic algorithm to evolve the network, its connections and activation functions (per node).

It got quite good.

Perhaps not that surprising that one can encode a lot in the graph structure itself, and given how DNA is quite good at dictating complex organs, I don't see why something similar couldn't be at work for insects and mammals alike. After all, nobody taught the newborn kangaroo babies how to crawl up into the pouch of their mom either.

Probably this (awesome) paper: https://proceedings.neurips.cc/paper/2019/hash/e98741479a7b9...

Yep, can't recall if someone else implemented it or if it was official repo tho. Was a car driving over obstacles in 2D IIRC.

Carl Jung held this view too. And he knew his teacher Freud disagreed. For instance a Carl Jung thought a humans fear of spiders in innate, and not learnt.

The same as how a bird can be scared on snakes. Interestingly native birds in Australia aren’t scared of humans.

> For instance a Carl Jung thought a humans fear of spiders in innate, and not learnt.

I don't have fear of spiders (was surprised by how many people do fear spiders). When I was young, I "hosted" a bunch of them under my table. Thinking of it, I think most people have their fears transmitted through their mothers. My mom certainly didn't fear spiders but she hated their "dirty" webs.

For example I do have a strong reaction to cockroaches till this day; a harmless insect that my mother fear. I can't even look at the beast. My cousin do not and conversely do fear spiders.

Curiously, though, we don't really know it is coded in DNA.

Michael Levin has been discovering lately that biology is more complicated than we thought. His YT videos are mind-boggling. Flatworm genetics are a dog's breakfast that the cells work around by means nobody understands.

His talks are some of the most fascinating I’ve ever heard. Nick Lane is a close second.

Michael Levin's work is awesome

What do you mean by this specifically? What claims does he make? It's hard to respond to this just based on what you wrote. I don't think biologists have claimed that biology isn't complicated, and I don't think anyone claims that DNA isn't complicated either. Unless you are referring to genetics? That is still caused by DNA, just not by gene expression.

They don't just build one. You may mean they don't learn from observation of others or from their mother. They may still acquire a sense of a better trapdoor and make better trap doors by experience. I can't think of an experiment to test the theory, but it might be around qualities like "thickness" or "wind resistance" if you could make a condition of test which put the trapdoor they do un-biassed under specific stresses.

Still, even if that's true, it does not refute the point. The knowledge necessary to build the very first burrow or web, or trapdoor (which will be definitely more than good enough otherwise spiders would no longer exist) is hardcoded, not externally acquired.

Isn't virtually all behavior of non-humans driven by instinct? Our learned behavior seems way more unlikely and amazing.

Are you saying that humans are the only animal capable of learning?

No, learning from parents is common for the young of most mammals and many birds, at least.

Primates are on another level compared to most other animals, and humans are definitely outliers, but learned behaviour is far from unique to humans.

I agree. An individual may develop improvement over time (43 years!) but there's no evidence of transmission across generations. Still, ability to refine might be heritable?

> the Tabula Rasa theory is wrong

Good book about precisely this (and in agreement with your statement) is cognitive psychologist Steven Pinker's 2002 book The Blank Slate: The Modern Denial of Human Nature.

https://en.wikipedia.org/wiki/The_Blank_Slate

I’ve thought about differing levels of innate behaviors in different species as being somewhat akin to different sorts of programming languages. So dogs and cats are more akin to, say, Pascal¹ which provides a high level of functionality at the language level, but are limited in capability² while humans are more like C where there’s less functionality out of the box, but the ability to accomplish unique tasks is less bounded.

⸻

1. This is a theory I came up with as a high school student in the 80s so my choices of programming languages reflect that.

2. Insert obligatory Turing-completeness disclaimer here.

I think the biggest issues involved around knowledge/skills/intelligence being potentially linked to genetics is that it often draws the eugenics proponents into the discussion and it's obvious how that conversation ends.

Sometimes I feel like my fear of spiders is baked into my DNA. I think I found some evidence to support this, can't remember where though...

It has been well argued that "The Blank Slate" argument is wrong, or at least more complex and subtle than that.

In Pinker's book The Blank Slate" in particular.

We are born blank of our mother tongue, but are wired to acquire one. Not wired with the skill itself, but with the propensity to acquire it. Both nature and nurture are deeply important to the outcome.

It reminds me of how Noam Chomsky speaks about language / language learning being something inate in humans.

the children of time series explores this concept in spiders. What happens if you were able pass down all of your knowledge into your children through genetics and how would that change society.

i think we are going to find way more comes from dna than we thought. Physiognomy is the science of determining your personality (and other traits) from the way your face looks. this was cast aside as pseudoscience like palm reading, until recently ai facial recognition found its real

https://en.m.wikipedia.org/wiki/Physiognomy

Interesting-- I wonder if criminal behavior is embedded in DNA...

And now we're heading towards eugenics once again. Amazing how quickly these discussions turn that direction.

> What amazes me about spiders is their complex unlearned behavior. No trapdoor spider learns how to build a trap. No orb weaver learns how to spin a web. Trap/web building behavior is entirely programmed in their DNA.

Do we really have evidence for this rather than observation and communication from parents to offspring. I'm deeply skeptical.

I have no idea about this, but, how hard would it be in theory to test? Just remove a spider from any other spiders and see if they develop that behavior.

I don't know if the parent comment represented the expert consensus correctly, but if they did, I wouldn't automatically be skeptical of it.

we do https://en.wikipedia.org/wiki/Nature_versus_nurture#Intellig...

That evidence is compelling and well-reviewed. But that evidence is also countered with other evidence: the debate isn't settled. But it's undoubtedly clear that a part of our intelligence and other personality traits are inherited: coded into our DNA. What's not settled, is how much.

Spiders that hatch without their parents around build the same webs as any other spider from the same species. This behaviour is really not something that leaves any space for being sceptical.

Is it too much to believe that both hardware (built-in) and software (flexible) parts are employed in nature?

Reply on news.ycombinator.com