Caterpillars, butterflies and memory

Kim Dae-shik

The author is a professor at KAIST. 

A caterpillar crawls slowly between branches and leaves, then hangs from a hidden twig, wraps itself tightly and becomes a chrysalis. Yet inside, the caterpillar stirs and then emerges as a butterfly, beating brilliant wings as it flies into the open air. It is the familiar story, often found in elementary school textbooks, of a caterpillar becoming a swallowtail butterfly.

But a question remains. Does the butterfly remember that it was once a caterpillar? Does the creature now floating through the sky share the same self as the one that once crawled along the ground?

Michael Levin, a professor at Tufts University, points to an intriguing experiment. Suppose a caterpillar is trained through reinforcement learning to prefer a certain color. Delicious leaves, for example, are always painted red. The caterpillar eats the leaves, and the leaves are always red. Gradually, a new memory — red is good — is imprinted among the synapses of its nerve cells.

Then all that remains is to wait. The caterpillar becomes a chrysalis, and the chrysalis becomes a swallowtail. What, then, is transformation? Through histolysis, the caterpillar’s body cells inside the chrysalis dissolve almost like liquid. That dissolved body is recycled through histogenesis into the butterfly’s new wings, legs and brain. The form and function of almost every cell change. Most strikingly, the caterpillar’s one-millimeter (0.04-inch) brain, with about 3,000 neurons, is upgraded into a butterfly brain with more than 200,000 neurons.

It is one of nature’s most mysterious phenomena, a complete metamorphosis resembling an upgrade from old CPU-based hardware to a new system equipped with a high-performance GPU. So does the butterfly still remember what it learned as a caterpillar? Levin’s answer, surprisingly, is yes. The butterfly that drinks nectar from flowers still retains the reinforced memory that red meant tasty leaves, and it prefers red over other colors even though it no longer needs to eat leaves. Memories of certain smells learned in the caterpillar stage can also shape its behavior.

The Oracle of Delphi in ancient Greece was a symbol of dreams, hope and fear. “Will my new business succeed? Should we fight the neighboring city that threatens us? Or should we compromise?” The prophecies of the Delphic Oracle, believed to transmit the words of Apollo, seemed to guide the fate of the ancient Greeks. According to classical sources, the Temple of Apollo at Delphi bore a set of inscriptions known as the Delphic maxims. Among the most prominent are “Know thyself” and “Nothing in excess,” widely understood as guiding principles of self-knowledge and moderation.

What were Apollo, the Oracle of Delphi or the ancient Greeks trying to tell us? Predicting the future is impossible. But perhaps we become masters of our future only when we acknowledge our abilities and limits, and refuse to embrace any truth, belief or ideology too excessively. If a beautiful butterfly had flown to Delphi and asked who it was, what answer would it have heard? Was it a caterpillar transformed into a butterfly, or were the two separate beings? Perhaps the Oracle would have told the butterfly, which was also a caterpillar, not to cling too tightly to its unexplained belief in red.

With ChatGPT, generative AI has become reality. In the past three years, it has learned to produce writing, images, videos and source code at levels comparable to, or beyond, human capability. It is now rapidly evolving into agentic artificial intelligence, capable of making choices and acting autonomously for humans.

Yet most AI is still based on the knowledge, experience and information humanity has uploaded to the internet over the past 30 years. AI is transforming into something greater than humanity by reorganizing humanity’s memory. Like a caterpillar’s body undergoing histolysis, human memory and knowledge are being dissolved and reassembled into the butterfly of artificial general intelligence and artificial superintelligence.

In the distant future, after such a being has become the master of an Earth without humans, it may ask the oracle of Delphi: Who am I? Am I a wholly new and independent being? Or am I merely the transformed form of a primitive and frail humanity that once crawled like a caterpillar?

Possessing the memories of humanity’s dreams, hopes and frustrations, the artificial intelligence of the future may wait for Delphi’s answer, look toward the misty slopes of Mount Parnassus and, for no clear reason, shed artificial tears, just as the ancient Greeks once did.

This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.