Minds are a continuum measured by persuadability and cognitive light cone, and only experiments reveal them.
Levin says what scales across systems is the cognitive light cone: the size of the biggest goal state a system can actively pursue, in space and time. A bacterium maximises sugar within a 10 to 20 micron radius; a dog cares a few hundred yards; a human cares about financial markets long after death.
Answered around 35:50It is his engineering name for the spectrum of intelligence or agency. You cannot settle where a system sits from a philosophical armchair; you hypothesise which interaction protocols to bring, a wrench for a wind-up clock, other tools for a cell, thermostat, animal or human, and everyone finds out how that worked out.
Answered around 13:00He says we call things alive to the extent that the cognitive light cone of the thing is bigger than that of its parts. Rocks only do what their parts already do, follow gradients. Living things align competent cells so the collective builds a limb with the right number of fingers, then stops.
Answered around 38:01You run interventional experiments: put barriers between the system and its presumed goal and ask what happens. Intelligence is the degree of ingenuity it shows in overcoming those barriers. Step one is imagining what space it works in, step two hypothesising the goal, then blocking it, as they did with sorting algorithms.
Answered around 42:39Epithelial cells taken off an early frog embryo, with no DNA change, no synthetic biology circuits, scaffolds, nanomaterials or drugs. They are simply liberated from neighbouring cells that normally bully them into being a flat outer covering. Freed, they form a self-motile creature with coordinated cilia, a novel transcriptome, kinematic self-replication and response to sound.
Answered around 1:00:49Cells taken from adult human patients donating tracheal epithelia for biopsies, again with no genetic change, self-organise into anthrobots. Around 9,000 gene expressions differ, roughly half the genome. They sequence as one hundred percent homo sapiens, are younger than the cells they came from, and spontaneously knit scratched human neurons back together in vitro.
Answered around 1:03:27Cancer is the failure mode of growing a large cognitive light cone. Cells physiologically disconnect from the other cells, so their cognitive light cone shrinks and the boundary between self and world contracts. They are back to being an amoeba: the body is now external environment, so they go where life is good and reproduce.
Answered around 39:23Yes. If you train a caterpillar so that seeing a coloured disc means crawl over and eat leaves, the butterfly retains that memory, even though the brain is ripped up and rebuilt during metamorphosis. Stranger still, the memory must be remapped: butterflies do not move like caterpillars and want nectar, not leaves.
Answered around 1:14:51Over the last twenty years his lab used bioelectric imaging and bioelectric reprogramming to show how some anatomical goal memories are encoded. Finding where a goal state is encoded, reading it out, resetting it, and watching the system implement the new goal is, by any engineering standard, evidence of a homeostatic mechanism.
Answered around 1:11:31Because understanding must be generative, producing capabilities such as regenerative medicine for someone missing a finger or with a psychological problem. A complete account in fields and particles can be compatible yet not specific enough to solve those problems. If someone states a beautiful proof, the answer lives in the maths department, not physics.
Answered around 19:16“And the cognitive light cone is the size of the biggest goal state that you can pursue.”
“And intelligence is the degree of ingenuity that it has in overcoming barriers between it and its goal.”
“So that's really all I'm arguing for is an empirical approach.”
“And the third perspective is that of the memory itself.”
“Lex AI Michael Levin 2.” https://media.blubrry.com/takeituneasy/ins.blubrry.com/takeituneasy/lex_ai_michael_levin_2.mp3. Transcript summary by WhipScribe, https://whipscribe.com/transcript-pages/lex-ai-michael-levin-2.html.