What we—and AI—can learn from nature's intelligence

Episode Summary

Executive Summary: This episode explores "natural intelligence"—remarkable problem-solving in plants, insects, immune systems, and animals—and how it can inspire better AI, low-power computing, health understanding, and conservation tools. From Venus fly traps that count to dragonflies that intercept prey, the show argues that intelligence is not uniquely human and is increasingly being decoded and modeled for practical use.

Main Topics: Plant intelligence and computation (Priority: 5/5): Greg Gage demonstrates that Venus fly traps count trigger touches before snapping shut, and that mimosa and peapod experiments show plants can process information and make flexible growth decisions. Definition of intelligence beyond humans (Priority: 5/5): The episode reframes intelligence as the ability to reach goals with available resources, suggesting that cells, plants, and simple organisms exhibit forms of intelligence worth studying. Bio-inspired AI and low-power computing (Priority: 5/5): Frances Chance explains how dragonfly hunting, dung beetle navigation, and ant pathfinding can inform efficient algorithms, chips, drones, and navigation systems that use far less energy. The immune system as a behavioral intelligence (Priority: 4/5): Keely Muscatel describes how cytokines drive sickness behaviors, mood changes, and social withdrawal, showing that the immune system actively shapes behavior to aid recovery. AI-enabled interspecies communication (Priority: 5/5): Karen Bakker shows how bioacoustics and machine learning can decode bat, orca, whale, bee, coral, and plant sounds, potentially enabling conservation and new forms of interspecies understanding. Ethics and conservation applications (Priority: 4/5): The episode considers both the promise and ethical risks of translating animal communication, while emphasizing practical benefits like whale-ship collision prevention and coral reef restoration.

Key Arguments: Intelligence should be understood as the ability to get to what you want with what you have, not just as human-like reasoning. Venus fly traps are not merely reactive plants; they compute by counting successive touches and only closing when a real meal is likely. Plants and slime molds process information without brains, implying that computation can be distributed across cells rather than centralized in a nervous system. Dragonflies solve fast interception problems in a few neural steps, offering a model for efficient AI and low-power hardware. Biology can inspire technology that is faster, smaller, and dramatically more energy efficient than current computing systems. Inflammation is not just a symptom of infection; it changes mood and social behavior in adaptive ways that encourage rest and recovery. Chronic inflammation can become harmful when stress, sleep loss, and lifestyle factors keep the immune system activated over time. Machine learning can decode hidden acoustic signals in nature, revealing complex communication systems in species we used to underestimate. Bioacoustics can produce direct conservation wins, such as protecting right whales from ship strikes and helping restore coral reef soundscapes. Any effort to translate non-human communication should be guided by strong ethical guardrails, since eavesdropping on other species may be intrusive.

Data Points: Venus fly trap closing threshold: 2 touches within about 20 seconds - Greg Gage explains the plant counts successive trigger-hair touches before snapping shut Trap reopen time: 24 to 48 hours - If no fly is inside, the Venus fly trap takes this long to reopen Fly-trap lifespan of trap function: Only a handful of closures - Each trap can only open and close a limited number of times before dying Dragonfly prey capture success rate: Up to 95% - Frances Chance describes how successful dragonflies are at hunting chosen prey Dragonfly response latency: About 50 milliseconds - Time from prey movement to dragonfly response Neuron integration time: About 10 milliseconds - Used to explain why only a few sequential neural steps are possible in dragonfly interception Estimated dragonfly neural layers needed: At most 4 layers - Based on the 50 ms response window and 10 ms per computational step Human brain power use: About 20 watts - Used to compare biological efficiency with future brain-inspired computers Right whale ship-strike deaths in protected zone: 0 since program launch - Bioacoustic whale-tracking system alerts ships to slow or move away Right whale population mentioned: Less than 400 - Karen Bakker notes the community protected by whale lanes is very small Watershed population served by whale protection: Tens of millions of people - Shows scale of human activity managed alongside whale conservation Plant health classification accuracy: About 70% - An algorithm listening to plant sounds can distinguish healthy, dehydrated, or injured plants

Pivotal Quotes: "The coolest thing is that the plant can count." — Greg Gage: Used to introduce the Venus fly trap experiment and challenge assumptions about plant cognition "Intelligence is being able to get to what you want given what you have." — Greg Gage: A simple definition offered to broaden the concept of intelligence across living systems "When we lose species, we lose voices. When we lose landscapes, we also lose soundscapes." — Karen Bakker: A closing reflection on why bioacoustics matters for conservation and empathy

Implications: Listeners are encouraged to see intelligence as widespread in nature and to value bio-inspired science. The ideas point toward more efficient AI, better health models, and powerful conservation tools, while raising ethical questions about using technology to decode animal communication.