What if Eye...?
eyes.mit.edu
eyes.mit.edu
He was (and hope he still is!) a great person, who taught me being curious and insightful in the computer science field. Taught me about Donald Knuth, TeX and many more things.
I can’t conceptualise this. How? Surely this is the same as humans where if you had a big enough uniformly lit object in front of you, you couldn’t “see” it?
This seems like a very basic condition, and that going from 0 to 'just a single sheet of light sensitive cells' is almost nothing. But of course that is not the case. Before that you would need (+++):
1) Photoreceptor Proteins 2) Functional nervous system or signal processing pathways 3) A machinery to translate the absorption of light into an electrical signal 4) A system for coordinating these signal with other parts of the organism . . .
Iirc, some snakes have some heat sensing cells arranged in pits to give them a directional heat sense. If you did that with visible light instead of IR those pits would be what halfway to an eye would look like
Like light sensitive cells these would need a similar 1 - 4 +++, so the point would really be the same.
A cell sensing light can use the same type of nerves etc as other sensory cells, so there is no need to explain how a cell sensing light + a nerve + a central nervous system evolves in one step.
If you wonder how a nerve could evolve for example, that is a different question. If you have two cells and they can exchange chemical signals, which is useful in itself, you have the start of a nerve.
If you're going to make that objection, you might as well head straight up the chain and aim for the biggest example of all: how does the first self-replicating molecule appear?
If we're going to study an evolutionary process, we have to pick a starting point. "How an eye evolves from a single sheet of light sensitive cells" is particularly relevant because creationists like to claim that "half an eye isn't useful, so it couldn't have evolved".
"How does a single cell evolve light sensitivity" is not a particularly hot button topic. Creationists don't drone on about plants.
The other flavor are the ones saying "wait a second microbiology is really complex, I'm not sure the standard evolutionary process cuts it here."
Saying, "well let's just assume microbiology isn't a problem", isn't going to be compelling to them.
It is an interesting endeavor, but with respect to the whole creationism vs evolution discourse it feels like a complete waste of time.
For exmaple, for #3, you say electrical signal. But cells normally use chemical, not electrical signals.
For #1- nope, just need an enzyme or other thing that makes a pigment molecule, and something that can detect a chemical (re-use of an existing protein)
For #2- no nervous system, and cells already have many internal signal processing systems.
For #4- those already existed.
Biology mostly just copies and re-uses systems that already exist. It's still incredible there was a path of mostly random events that led to fully formed eyes, though.
You portray such a +++ animal as simple and obvious, but indeed in fact you would need (++++++):
1) adherins and cadherins to bind this sheet together 2) an organelle to produce and regulate the cellular proteins 3) a method to encode those proteins 4) a method to replicate and propagate that encoding so that new proteins can be encoded 5) start and stop codons and a method of duplication or recombination errors to allow for new copies of protein encodings to be created separately from old copies...
I think we're quite far off if you want to model full organism complexity. But if you want to answer a research question you can model simpler versions today. Like this research team did around how vision evolves.
By the way– As of recently, we were able to model C Elegans (flatworm) in 3D with all neurons and neurotransmitters. It reacted to virtual stimuli just like a real worm (https://www.nature.com/articles/s43588-024-00738-w). So single-organisms is already possible. But the evolution of these entities in 3D will take us a bit more time is my guess:
Also I'm not a CS person, just an enthusiast.
Can these two things interface yet is the question (virtual to real bidirectional interface), especially since you are suggesting we have a clone of it.
If we do this experiment, what does that say about something doing an experiment on humans (humans control humans inside video games).
Sounds insane right? Why would we ever do this? Well … we have this perfect digital clone of a flatworm, what else are we going to do? There’s a lot of evidence that humans would absolutely go down this rabbit hole until it’s logical conclusion.
One word: Teledildonics
Anyway, the flatworm that is born into such an experiment would never know, or it would just be useless to know. :shrugs:
Over time this allows fish to develop basic behavior such as searching for food, navigation a maze, etc.
The only other 'useful' gene right now is around herbivore/carnivore digestion (0 to 1), which allows them to extract more energy from either meat or plant-food. Most of the time they actually develop specific behavior according to this gene.
I don't really code in what offspring need to do beyond having slight variations to both factors described above, it kind of evolves randomly into more complexity (neural net + behaviors).
Also importantly– I need to program an energy decay system and death if they run out. So basically: Energy source, energy decay and evolving neural nets that can give an organism the possibility to survive and evolve if they get more energy. And voila– Life emerges.
Working on plants now, and again simple rules: Neural nets in the plants to mimic evolution of complex biological systems that evolve from generation to generation. And a light-based energy source and light-based energy capture system (leaves). My current (preliminary) experiments show that the plants start to look like trees over time to maximize energy capturing compared to competing plants.
Looking to publish this once I have it a bit more refined.
One more question, sorry my knowledge of ML is not so profound, does NEAT algorithm mimic how natural selection works? And how.
Natural selection in this sim just happens by itself, there is a limited amount of food and only the best adapted ones survive. So the best performing neural networks duplicate themselves and create small variations of themselves. This part is not connected to the NEAT algorithm, I've just seen that NEAT performs particularly good vs more fixed-structure neural networks.
Depends how accurate you want your model. Pie in the sky thinking, we need quantum computing before I can imagine these kinds of simulations making sense
https://www.microsoft.com/en-us/research/blog/mattersim-a-de...
More generally, ML is good at approximating many NP-hard problems efficiently, so I wonder if it will be a more practical alternative to quantum computing for things like molecular simulation.
And just because, for example, organisms that live in cold climates generally have thick hair, having thick hair doesn't imply cold climates. Some mammals fill niches filled by birds in other ecosystem, or by fish in others. Likewise in New Zealand they have birds filling niches that in other ecosystems are filled by mammals.
I'm not sure biology can be a purely inductive science.
Same goes for many of Ted Chiang's stories!
Red eyes are also caused by a particularity of vertebrate eyes, which have their blood supply in front of the retina.
This is also the reason why you can sometimes see moving dots when looking at a bright, blue-coloured thing like the sky: your eye sees the whole capillaries filled with red blood cells, but the brain processes them out because they're always there. The large white blood cells then appear as "less red" dots in the processed-out streams of red, and the brain interprets them as bright dots.
Insects have none of this, they don't have blood or blood vessels but a transparent haemolymph that doesn't really circulate through a complete circulatory system like us.
So there really is nothing in front of the insects' light sensors to reflect light.
Upon inspection, it was the light reflecting off the eyes of spiders. Dozens and dozens of spiders everywhere in the grasses. Never could get a picture.
So light does reflect from compound eyes, but in ‘silver’.
Here, the takeaway is that the emergence of two different types of eye – compound and camera-like eyes – can be modelled by a set of 3 specific tasks, in combination with a minimal set of anatomical knobs and switches. Then it might actually be _informative_ to compare and contrast the clear evidence from the model, and see how these explanations compare to the less conclusive ones we can draw from the methods of evo-bio.
(A good analogy would be to look at how gates, latches, and clocks can alone account for the "emergence" of modern superscalar microarchitectures, without having to resort to modelling the analog madness of pushing high frequencies through physical circuits, for example.)
---
Website seems broken (repo deleted?), had to go here[2] instead.
And here is the paper[3] the website is about.
> [ mTOR, Muller glia in Zebrafish, ]
From "Reactivating Dormant Cells in the Retina Brings New Hope for Vision Regeneration" (2023) https://neurosciencenews.com/vision-restoration-genetic-2318... :
> “What’s interesting is that these Müller cells are known to reactivate and regenerate retina in fish,” she said. “But in mammals, including humans, they don’t normally do so, not after injury or disease. And we don’t yet fully understand why.”
- The mechanism to interpret the light data signal has to be in-step with the evolution of the eye. Getting light data without a brain evolving at the same time to interpret it is evolutionary recessive, i.e. a useless function. I.e. a real evolution would be more like "cat /dev/urandom > output.html", not a controlled ecosystem with a clear penalty-reward system.
- In nature, there is no 1:1 "reward / selection function" like in this simulation. In the computer, this "motivation factor" is externally given, so that the next generation is rewarded and selected out, in reality, there is no rule as to what is and isn't "better" or "fitter" or "more attractive to the other gender" (not like CS nerds would know). Sure, an organism can consume food, but beyond a certain point that wouldn't make the organism just "fat", not stronger. So there also need to be environmental mutations happening at the same time, that reinforce "more food = better evolved".
- There has to be a way for the animal to be so dominant, that the connection between light data and food can be genetically passed on and will not be associated with bad artifacts (see ChatGPT hallucinations for examples of "accidental bad artifacts in evolution" - and that "evolution" has millions of man-hours, money and R&D behind it).
- By the rule of "survival of the fittest", the next generation mutation has to be (in one single step) such a significant improvement over the last one that it won't be selected out again by recessive selection or dilution inside of the gene pool.
- The gene has to be active within 150 subsequent generations, without fail, cancer, recession and provide 150 times a dominant advantage, just to get a basic "eye" for 2D navigation with 10 light sensors. The minimum snail eye (pre-Cambrian) has 14.000 cells [1] (and a snail cannot see color).
- The real world is a 3D environment, which adds a monumental amount of complexity. Add to it the complexity of depth, color, shape, ...
- The mutation(s) have to happen either "at once" or be widespread (otherwise it's going to be like an Albino animal, i.e. some rare neutral mutation).
- All of this has to be done in an environment hostile to life in general (i.e. the edge of underwater vulcanoes, some primordial soup burning at several hundred degrees), all elements have to be at the right place, at the same time, etc. And be created out of nothing, of course.
While I do agree that it can be helpful for computer vision, computerized "evolution" is just adaptive statistical pattern matching, but it's absolutely nothing like real biology. It would be more realistic to just output "/dev/random > kernel-gen-xxx.iso" and then run it bare-metal, with no lab environment, no operating system, no programming language, no goal function, no selection / reward process, no debugging, etc.
Even Darwin had his problems with the eye. The reason I believe in God is not necessarily because I want to, but because evolution (not survival-of-the-fittest, but the "mutation creates information" aspect) requires far more faith and far more dogmas, which cannot be questioned for the sake of science. When I was in 8th grade biology, I took a stone from the schoolyard, put it on the teachers desk and said "alright, so this is a human if we wait 4 billion years". The teacher ignored me, but never told me I'm wrong.
Spoiler warning: Darwin didn't really have any problems with the eye. That's just something creationists say.
Darwin's understanding of e.g. Heredity via Pangenesis turned out to be wrong, so it is not like just holding up a copy of 'On the Origin of Species' as the final judge of "origin reasoning" will take use very far.
The mechanism wasn't understood at the time yea. That's not really relevant, and not really covered in Origin of Species anyway.
Video codecs are good optical illusions too.
On a somewhat related note, due to a head injury I suffered now many years ago, I started developing small-ish blind spots "once in a while" that remain anywhere from a few minutes up to a few months.
The spots are very noticeable when they first appear, grabbing the attention all the time. The ones that persist long tend to "disappear" when my brain filters out the broken (?) signal of where the blind spot appeared, and the spot mainly becomes noticeable again if it hides or interrupts a known pattern that I'm looking at.
It's as if the brain fills the spot with the average color around it (blue sky - blue spot, white wall - white spot etc), which works well for single color surfaces and such but not great for repeating and predictable patterns. And the hiding "lags" which means if I quickly shift between colors then the spot will momentarily be visible as the old color shows up on the new color.
So the brain does "imagine" things, but when it does, it isn't perfect.