AI bulls believe Artificial General Intelligence or AGI is around the corner. AGI refers to the form of AI which beats human cognitive capabilities across any domain. Whilst AI is already beating humans at certain tasks, the comparison of machine intelligence with human intelligence has sparked a debate on what defines intelligence or cognition in the natural world itself. Research showing single celled organisms that go beyond just survival and reproduction without the need for elaborate neural networks is the subject matter of this brilliant piece by Samanth Subramanian, who is known to tackle mainstream subjects from an off-beat path, combine it with deep research and great prose. He focuses on scientists’ research on Slime moulds or Physarum, starting with that of Japanese scientist, Nagayaki: “In 2000, his astonishing research began to surface in the media. “Slime mould solves maze puzzle,” one headline ran, summarising an experiment in which Nakagaki induced a Physarum to find the shortest route through a maze in search of food….In 2010, the New York Times wrote about Nakagaki’s most famous experiment – the canonical study that scientists offhandedly call “the Tokyo railway one”. Nakagaki placed oat flakes to resemble the map of Tokyo and 35 of its satellite towns. Then he let the Physarum loose from the centre of “Tokyo”. From that one metropolitan oat, the mould spread outwards seeking the others, stretching towards them with its tubules. Then, putting out more branches, it linked up with itself, from one suburban oat flake to the next.
In doing all this, the mould displayed such a verve for efficiency – no wasted motion, an optimal pursuit of its food – that it ended up nearly retracing the route map of the greater Tokyo area’s railway system. Nearly, Nakagaki said – because, in fact, free of human bias or politicking, Physarum may have improved the design of the network. Later, when Nakagaki twice won the Ig Nobel – the prize for research that is both amusing and thought-provoking – the first award cited his maze paper in the field of “cognitive science”, but the second, cheekily, hailed the Tokyo railway study as a breakthrough in “transportation planning”….
…Which poses a mystery: how is Physarum, this cell without even a semblance of a brain, capable of it all – the decision-making, the memory, the processing of information?”
Subramanian connects the dots based on a scientists’ description of Physarum’s behavior: ““It grows across all of the empty spaces, and then withdraws from everything but the shortest route,” Pringle said. The way she described it put me in mind of large language models and their techniques of brute pattern-matching – an incomplete simulation of human thought that is described increasingly as intelligence.”
He then gives us a history of our understanding of intelligence: “In deciding what intelligence is, humans are easily influenced by the religions, politics and technologies that surround us. It’s been a wild ride. In the 17th century, Descartes believed all non-human animals were automata, lacking the immaterial mind that gives humans thought and reason. That view held even after Charles Darwin proved earthworms could learn and described the root tip of a growing plant as an analogue to the human brain. In both cases, he didn’t hesitate to use the word “intelligence”. Through the first half of the 20th century, the rise of behaviourism led scientists to disregard internal processes of thought altogether, to focus on an organism’s actions and reflexes only. For a while, even asking questions about cognition was frowned upon.
More tumult followed. Midway through the last century, computers became the new point of comparison for the mind, and it followed that humans too, like their new machines, thought by building internal models and systems of mathematical rules. This “cognitive revolution” only extended to organisms with brains – with neurons, the wetware version of the microprocessor. Then, in the 1990s, a roboticist named Rodney Brooks built a series of small robots that could trundle through the halls of MIT, dodging obstacles or collecting soda cans, all without constructing any mental models or simulations. This was a revolution in its own way – but as a philosopher named Fred Keijzer told me, even this was unsatisfying, because it failed to approximate the sheer complexity of, say, a jellyfish moving in the ocean. “Cognitive science is still at the stage that chemistry was before the periodic table,” he said. “It’s a mess.”
In this choppy way, we’ve come into the AI age, which exerts its own perverse pressures. Silicon Valley loves the biogenic camp – loves any move to categorise intelligence as a data and computation problem, since that would validate the I in AI too. No better way, one scientist remarked sardonically, to treat humans as machines than by starting to treat machines as humans. But placing humans and machines on a continuum of intelligence doesn’t in itself feel like a problem to Michael Levin, who has argued extensively that even single cells can learn and solve problems. “For Michael,” one scientist told me, “it’s just cognition all the way down.””
Subramanian lays out the debate on cognition in nature brilliantly connecting it with the current developments in AI.
If you want to read our other published material, please visit https://marcellus.in/blog/
Note: The above material is neither investment research, nor financial advice. Marcellus does not seek payment for or business from this publication in any shape or form. The information provided is intended for educational purposes only. Marcellus Investment Managers is regulated by the Securities and Exchange Board of India (SEBI) and is also an FME (Non-Retail) with the International Financial Services Centres Authority (IFSCA) as a provider of Portfolio Management Services. Additionally, Marcellus is also registered with US Securities and Exchange Commission (“US SEC”) as an Investment Advisor.
