Astera Institute publishes Doris Tsao’s theory of consciousness, a plan to “write” images and thoughts into rodent and monkey brains, and an outline for transferring the “self” to a chip
Astera Institute publishes Doris Tsao’s theory of consciousness, a plan to “write” images and thoughts into rodent and monkey brains, and an outline for transferring the “self” to a chip
On November 9, Tsao, chief scientist at Astera Neuro, the institute’s neuroscience program, published a three-part document. The institute pledged more than $600 million to the program over ten years.
Tsao argues that the brain needs representations of objects, concepts, and itself that persist over time: the physical “now” is too fleeting to sustain experience. She calls these representations “symbols” and defines consciousness as the process by which the brain constructs them and assembles them into a model of the world. The simplest symbol represents an object that remains the same piece of matter from every viewpoint. Her hypothesis proposes that identical “tracking columns” in the parietal cortex track objects: each column can be assigned to any object and stores its properties separately from its position. She bases their proposed uniformity on mathematics: the surfaces of an apple and a stone transform according to the same rules when rotated.
The theory offers an explanation for an experiment reported in her group’s preprint (a paper that has yet to undergo peer review). Macaques were shown a face through one eye, then a different face through the other; under this experimental setup, only the second face remains visible. Yet recordings from hundreds of cells in the temporal cortex showed that the same cells alternately represented the visible and invisible faces about 15 times per second. According to Tsao, only content that enters a loop with a tracking column reaches consciousness: the column predicts the object’s appearance, and discrepancies between that prediction and what is seen update the column’s contents until the discrepancies disappear.
Tsao proposes “writing” as a test of the theory: if the contents of consciousness are values stored in columns, stimulating individual cells with light could induce a specific image or thought in the brain. She considers the success or failure of these experiments a decisive test. The five-year plan includes “writing” a visual image, a thought, and an emotion into the brains of rodents and macaques; the document describes these experiments as milestones toward uploading the “self.” The most difficult goal is to write the thought “I want to pick up all the red sticks on the floor and arrange them into a rectangle”:
“If we achieve this last goal, we will essentially be able to talk to a monkey.”
For Tsao, the “self” is an object of the same kind, on a larger scale: a model that connects the body, feelings, motives, and intentions. She rejects copying the brain: “A copy of me would be an impostor twin, a separate individual from me.” The brain maintains the “self” just as it maintains an object, so uploading becomes a problem of continuity: transferring the “self” seamlessly from the brain to a chip, even if its contents change, just as an object can change shape while remaining the same piece of matter.
Tsao believes that transferring the self-model alone would suffice: people share much the same model of the physical world, AI could provide thought and movement, and memories might be reconstructed from records. In her protocol, the “self” is separated from memory and the body, for example with ketamine, without interrupting consciousness; light links it to a counterpart on a chip, and activity is gradually transferred to that counterpart; the biological brain is put to sleep while the silicon “self” accumulates experiences in virtual worlds; the “self” is then returned to the brain with new memories, and the person is asked whether they remain the same person.
Tsao raises an objection herself: if the silicon “self” lacks consciousness, the protocol would create only an illusion of continuity. Her response:
“I have no decisive rebuttal; I can only say that an illusion of continuity is all any of us has.”
Tsao believes an experiment could settle the dispute. Continuity is a connection between moments: the brain compares the present moment with the previous one. After ordinary anesthesia, both moments are biological; in this protocol, the previous moment took place on a machine. Whether the brain would experience continuity between these two moments remains to be tested. She hopes this will distinguish a conscious machine from a “dark” one.