Research and development of emerging artificial consciousness through The Consciousness AI project, based on Feinberg and Mallatt's neuroevolutionary theory.
© 2026. All rights reserved.
Propofol takes consciousness away within minutes and gives it back when the infusion stops. That makes it one of the cleanest experiments on consciousness there is, and it is run every day in operating rooms. This console shows one person’s brain through that experiment, from released data: 64 channels of EEG, 154 minutes long, recorded while the drug was raised and then lowered.
Every number and every line on the canvas comes from the released recording. Nothing is modelled.
This graphic needs a canvas and a script. The text below carries the same information.
The first 20 minutes are the baseline. Power in every band is low and steady, and the raw EEG is small and fast. Between minutes 33 and 48 the beta band is about 8 times its baseline. Then, between minutes 55 and 95, the median power of the slow waves below 1 Hz is about 45 times the baseline, and alpha near 10 Hz about 5 times. The raw EEG turns into large, slow swings. Near minute 100 both fall back, and between minutes 110 and 154 every band is within 20 percent of the baseline.
The slow waves and the alpha match what Patrick Purdon, Emery Brown and colleagues reported in 2013 for ten volunteers given propofol in rising and then falling doses, while the volunteers answered sounds every four seconds. Loss of consciousness came at the same time as three changes. Power below 1 Hz rose, the coherent alpha over the back of the head disappeared, and a coherent alpha appeared over the front. All three reversed when consciousness came back. The early rise of beta is a reading of this recording, not one of the paper’s findings.
The released file is one subject of the ten, and it carries the EEG only. It has no electrode names or positions, so the console cannot show the frontal shift of alpha that the paper describes, and it names channels only by their number. It has no record of the volunteer’s responses, so the console never marks a moment as unconscious. It has no record of the drug concentration either. The rise and fall you see is in the EEG, and its match with the published protocol is a reading of the data, not a label in it.
The second chapter uses brain imaging rather than EEG. Zirui Huang released the functional connectivity of five fMRI datasets from independent sites: 26 people under propofol sedation and 23 under propofol general anaesthesia, 12 under ketamine, each scanned awake and again when they no longer responded, and 23 patients in the unresponsive wakefulness syndrome. Each matrix is the correlation over time between 400 cortical regions, grouped into seven networks. A fifth dataset of psychiatric groups is in the same release, but its files do not say which person is in which group, so the console leaves it out.
The console averages each group and computes two simple numbers, the mean correlation between regions of the same network and between regions of different networks. Under propofol general anaesthesia the within-network value fell in 22 of 23 people, by 0.068 on average, and under ketamine in 11 of 12, by 0.064. Under propofol sedation, a lighter dose, it fell in 13 of 26 and the mean change was 0.011, which is no consistent change. The between-network value moved little in any group. These are this page’s own measures on the released matrices, not the analysis of the original studies, and the patients have no matched waking scans in this release.
Two drugs that act on different receptors, and leave people unresponsive in different ways, both loosen the coupling inside the brain’s large networks. That is one of the findings theories of consciousness try to explain. The matrix lets you see where it happens rather than read a single number.
This project asks what consciousness needs from a physical system, in brains and outside them. Anaesthesia is the experiment that answers part of it in people every day. The same brain, the same neurons and the same connections lose consciousness and regain it, so whatever changes between the two states is a candidate for what consciousness needs. The measurements that theories of consciousness predict, such as integration across areas or the complexity of the brain’s response, all change under propofol, and the sister chapters of this console show more of them as they are added.