Hacker News Reader

Comment by ta20211004_1 | Hacker News Reader

ta20211004_1

I find the “super computer slime” stories that pop up every once and a while (including in academic contexts, occasionally), pretty annoying.

Usually they have fun examples of problems these things can solve, but we don’t actually think they’re doing something fundamentally different from a greedy algorithm or trial and error optimization? And they’re slow as slow can be. So, not that exciting?

View on HN

They're primarily exciting because they exist at a fringe of research and challenge prevailing biological dogma that specialized organs (specifically the animal kingdom's specialized organs, the nerves) are required for complex behaviors and intelligence. They are an extremely important research area for philosophy of mind and their study has only provided more and deeper mysteries.

Slime molds aren't ready-made solutions to practical problems. They're a signposted paradigm shift that indicates we need to re-evaluate some of our a prioris. Alongside advances in understanding complex behaviors in plants, slime molds represent one of the largest challenges to long-standing assumptions we've had for millenia. Unfortunately the response to recorded behaviors of adaptation, memory, learning, problem solving in things without neural structures has hilariously been to shift goal posts and underline said neural structures as a prerequisite to these things being meaningful. Silly things like "super computer slime" and pop science articles drown out the incredibly important experimental observations made about these beautiful organisms, this much I agree with.

Thanks for your perspective and insights. I'll take a more charitable view on this topic in the future. :)

Try this lens: dragonflies are one of nature's most incredible predators, with a 90-97% success rate of hunts. The precise data varies, but almost all more complex animals have a dramatically lower success rate. Well, so what? Harbour porpoises have a ~90% success rate, too.

It's interesting because dragonflies only have about 10,000 neurons, whereas harbour porpoises have like 14 and a half billion. Slime molds aren't meaningfully intelligent at all, but they can work out problems our primate brains had to work quite a long time to systematize. This suggests, to me at least, that dragonflies and slime molds are worthy of further consideration.

...that said, I try not to read the popsci articles about either.

Those are some interesting numbers, but I don’t understand the connection. Slime molds are exciting because dragonflies are better hunters than porpoises despite having a millionfold fewer neurons?

Both slime molds and dragonflies are amazing because they flaunt much of what we know/believe about intelligence.

Neurons are obviously excellent at handling and developing "intelligent" behavior, but it turns out they're not as necessary as we thought and there may be extreme efficiency gains possible (for specialized problems) over what traditional neural architecture indicates.

Except aren't slime molds on the extreme end of inefficiency, seeing how they solve problems by physically exploring the problem space? Being able to map the environment remotely, via senses like sight and hearing, and being able to solve problems computationally, "in mind space", are huge efficiency wins.

I was more referring to dragonflies' low neuron count with efficiency, but really efficiency is relative. You're describing speed, or efficiency with respect to time. We're able to do these things faster than a slime mold, but it take a lot more energy. Slime molds can be dropped on the floor and start solving problems, we have to build tools to hunt animals to roast over a fire we've built so we can sleep in a shelter we've built.

I'm just thinking in terms of energy expenditure on the margin. Humans need more energy to sustain themselves, but for a given cognitive task, I imagine they need to expend less energy than slime mold - being able to solve problems by gathering photons and sending electrons down a maze is much more energy-efficient than having to move or grow your body, made of matter, to solve the same problem by touch alone.

That's fair and may. I don't know what the caloric needs of a slime mold are. But this is completely off topic, tbh. The point is that they solve problems in a way that is alien to us, and alien intelligence is interesting. I may have given a poor explanation before, but efficiency is not the important part.

Dragonfly neurons encode an extremely good hunter / seeker missile algorithm but they can’t solve other types of problems. So it can’t be argued that dragonflies are generally intelligent. There is more evidence for bee intelligence.

Neither I nor the article argue for general intelligence in slime molds or dragonflies. Kind of a strange point for you to make, frankly.

maybe more complex animals don't have as high a success rate because capturing the prey they go after is a more complicated problem.

There are works using slime mold as logic gates. But i do not think any work has gotten past that to make a full on computer.

https://www.sciencedaily.com/releases/2014/03/140327100335.h...

https://www.sciencedirect.com/science/article/pii/S136970211...

edit: it is also used as a sensor:

https://techxplore.com/news/2022-12-scientists-smartwatch-po...

It's of the same vibe as "electricity solves a maze" or "water solves a maze", imo

No, it's not even remotely comparable. Neither electricity nor water changes propagation rules in response to stimuli and pre-emptively employs techniques to adapt to changing conditions.

Water's propagation rules for solving a maze are in fact dynamic and definitely not what you might expect. I bet if you wrote a sim for it you would get it wrong on your first shot.

Reply on news.ycombinator.com