- Recursion
- Posts
- Imagine learning without a brain. Pretty tough, right?
Imagine learning without a brain. Pretty tough, right?
What can development of artificial intelligence and robotics learn from jellyfish?
a school of jellyfish, water color, monochrome bluegrey background, simple composition, --ar 16:9
In short: Scientists have discovered that a species of jellyfish can learn (retain knowledge) about patterns in their environment and course correct based on that knowledge. What’s remarkable is that they’re doing so without a central nervous system, what we typically call a brain.
Deep dive: In a controlled environment, scientists replicated naturally occurring habitats in both murky and clear waters in cylindrical tanks. In the murky water tank, the jellyfish would initially bump against the wall because their simple “eye-brain” nerves could not see the obstacle in time. But in a matter of minutes, the combination of visual and mechanical stimuli (what it observed while bumping in to it) taught it to pulse rapidly when closing in on the edges, avoiding collision.
The scientists still need to understand whether and/or for how long this knowledge is retained plus how it is distributed between its four different nervous centres, that control its movement.
Context: This is of course not the first encounter of intelligence elsewhere in the animal kingdom. Invertebrates like octopuses are the most common example of advanced animal intelligence, but also corvids like ravens and dolphins have shown to exhibit cognitive abilities on par with or better than human cognition. However, all these animals have brains like we do, so that is where the novelty of the discovery comes from.
The scientists now hope to learn whether it is beneficial to apply similar principles of learning in the field of robotics and artificial intelligence. It wouldn’t be the first time naturally occurring phenomena had inspired AI, and likely not the last. For instance, the algorithms behind e.g. StableDiffusion (an open-source AI image generator) is inspired by diffusion of heat works universally and naturally occurring mould growth patterns are inspiring the generative design algorithms that enable companies like Airbus to construct airplane frame walls to be equally resistant while reducing weight.
The discovery adds increasing urgency to the identification and mapping of undiscovered species in the context of the 6th ongoing mass extinction event and the human-induced pressure on ecosystems. To give some perspective: A scientific review from 2011 estimated that 86 percent of land species and 91 percent of marine species have yet to be identified and described. What if there’s so much more out there that we can learn about the plurality of intelligence that is vanishing under our watch?
My thoughts: Our assumptions and biases around what intelligence is and and what it’s not are about to be challenged: human-made general artificial intelligence (AGI) is inching ever closer and our ability to search/discover extraterrestrial intelligence is improving with the emergence of reusable rockets and the wonders of the James Webb telescope.
Maybe we should adopt a more humble, curious and caring mindset when it comes to understanding and co-existing with other types of biological intelligence on this planet. Who knows, we might even discover insights about how to thrive beyond it.
Look out for: If we were able to replicate this type of intelligence in new, autonomous robotic systems, what should we put it to use for? What would it excel at? I am thinking automised plastic pollution cleaning systems for the ocean. Or Wall-E style space debris collection… in orbit!
What do you think?
💬 Well said
“While building AI to see what humans can see was the inspiration for computer vision, we should now be looking beyond this to building AI to see what humans can’t see.”
This quote from Fei-Fei Li in the recent report Generative AI: Perspectives from Stanford HAI hits the mark: Putting AI to work for humans and nature through enablement and augmentation of existing capabilities, not competing with them.
🧠 Worthwhile
A dilemma at the heart of the green transition. A beautiful and terrifying case from Canada’s wilderness: What do you do, when a massive amount of rare earth minerals - essential to the electrification and decarbonisation of modern society - rest beneath a significant carbon sink and indigenous land vital to preserving biodiversity? Wicked problems. The article is pay-walled, but you can read it here with my highlights.
LEGO dumps it’s biomaterial moonshot. It is obviously a shame for a company that size to archive ambition like that… but I can’t help but wonder what a similar-sized investment in the Replay Initiative - the circular offering that currently only is used for donations - could have had in terms of impact? Sometimes, business model innovation trumps material innovation.
Greening the web, one page at a time. The Worldwide Web Consortium (W3C) has released a draft of the first official Web Sustainability Guidelines. So much has happened in this space the last couple of years, but seeing it finally crystallised in a guideline on par with ones for accessibility and usability is amazing.
🌻 The bright side
Photo: Hawaii DLNR.
After the tragic wildfire rampage in Hawaii earlier this Summer, the beloved banyan tree in the heart of Lahaina, the hardest hit town on Maui, is showing signs of recovery, in part due to volunteers tending to it ever since the fire. Nurture and nourish: Those are meaningful activities we all can practice locally and globally.
Recursion is a regular publication of essays about the relationship between technology and the Earth’s biosphere. Once in a while, it might contain sci-fi short stories. Written by Morten Bjørn Hallkvist.
If you enjoyed reading this, feel free to subscribe below.