Natural Intelligence: What we and AI can learn from nature

Episode Summary

Executive Summary: This episode argues that intelligence is not uniquely human or even animal: plants, slime molds, immune systems, and animals all solve problems, adapt, and communicate in surprisingly sophisticated ways. Through live demos and TED talks, it shows how natural systems inspire low-power computing, better AI, medical understanding, and bioacoustic tools that can protect biodiversity and reveal hidden interspecies communication.

Main Topics: Plants as computational organisms (Priority: 5/5): Greg Gage demonstrates that Venus flytraps count trigger touches before closing, mimosa plants respond to touch with electrical signaling and movement, and even pea plants can adaptively learn environmental patterns. Redefining intelligence in living systems (Priority: 5/5): The episode broadens intelligence beyond brains, arguing that any system that can get what it wants with what it has—cells, plants, animals, immune systems—exhibits a form of intelligence. Dragonflies and brain-inspired computing (Priority: 5/5): Frances Chance explains how dragonflies intercept prey with rapid coordinate transformations, motivating neuromorphic AI and ultra-low-power hardware modeled on biological neural circuits. The immune system’s behavioral effects (Priority: 4/5): Keely Muscatel describes how cytokines create sickness behaviors like fatigue, low mood, and social withdrawal, while also sometimes increasing motivation to seek comfort from close others. Bioacoustics and non-human communication (Priority: 5/5): Karen Bakker shows how AI and bioacoustics decode hidden sounds from bats, whales, corals, plants, and bees, revealing complex interspecies communication and practical conservation uses. Ethics and ecological applications of listening to nature (Priority: 4/5): The episode considers whether translating animal communication is helpful or intrusive, and emphasizes bioacoustics for conservation, including coral restoration and whale-ship collision avoidance.

Key Arguments: Venus flytraps do a real computation: they count successive touches within a time window before snapping shut, conserving energy and reducing false alarms. Intelligence can be defined simply as the ability to get to what you want given what you have; by that standard, many organisms and even cells are intelligent. Dragonflies perform rapid, efficient prey interception using a small neural circuit, offering a model for faster and lower-power AI systems. Natural systems may inspire computers that require far less energy than today’s devices and data centers, potentially lowering carbon costs. The immune system is not just defensive; it also alters mood and behavior in adaptive ways to promote rest and recovery during illness. Inflammation can contribute to chronic stress and depression when prolonged, making body-brain feedback loops important to mental health. Bioacoustic AI is uncovering hidden communication in bats, orcas, peacocks, corals, bees, and plants, challenging the idea that silence means absence of signal. Listening to ecosystems can directly aid conservation, from restoring coral reefs with healthy soundscapes to protecting whales from ship strikes in real time.

Data Points: Venus flytrap trigger window: about 20 seconds - The plant closes only after repeated touches occur within roughly this time span. Venus flytrap reopening time: 24 to 48 hours - Used to explain why the trap conserves energy and avoids closing too early. Venus flytrap feed frequency: a handful of times per year - Shows the plant gets most energy from the sun and only supplements nutrients with insects. Mimosa response mechanism: water flushes out of cells - The plant moves by altering water pressure rather than muscles. Dragonfly prey-capture success: up to 95% - Describes how often dragonflies catch the prey they choose to pursue. Dragonfly response latency: about 50 milliseconds - The time from seeing prey movement to reacting, implying only a few neural computation steps. Neural step duration: about 10 milliseconds - Used to estimate that dragonflies have room for only about four sequential computational layers. Human brain power: about 20 watts - Presented as a benchmark for how energy-efficient brain-inspired computers might be. Plant health detection accuracy: about 70% - An algorithm listening to plant sounds can distinguish healthy, dehydrated, or injured states. Right whale strike outcome: no right whales killed by ship strikes in the protected zone since launch - Bioacoustics warns ships in real time to slow down or change course. Right whale population: less than 400 - Illustrates the urgency of acoustic protections for an endangered species. Orcas in the Salish Sea: only a few dozen left - Used to underscore the precarious status of local whale populations.

Pivotal Quotes: "The coolest thing is that the plant can count." — Greg Gage: Introduces the Venus flytrap demo and the idea that plants perform computation. "Being able to get to what you want given what you have." — Greg Gage: His compact definition of intelligence during the discussion of plants, cells, and animals. "When we lose species, we lose voices. When we lose landscapes, we also lose soundscapes." — Karen Bakker: A key line connecting bioacoustics, biodiversity loss, and conservation urgency.

Implications: The episode suggests intelligence should be understood as distributed across life, not centered on human brains. That shift could improve AI, medical care, conservation tech, and ethics around how we study and protect other species.