We think that neural activity creates subjective experience (“phenomenal consciousness”) but lack a satisfying, quantitative understanding of how that actually happens. Near the heart of this mystery is the binding problem, which I understand as having two main sub-questions:

1. The human brain is composed of billions of neurons over more than a thousand cubic centimeters, and neural activity relating to consciousness unfolds in time over tens of milliseconds, which we nevertheless experience as an integrated phenomenal “moment.” How does this spatiotemporally extended neural activity get “bound” together into a unified experience?
2. Our moments are structured into modes — vision, audition, somatosensation, proprioception, etc — that appear to be somewhat independent though which we experience jointly, and we always experience all of our own modes together and never, say, my vision and your hearing. How does this happen? Put differently, how is the “border drawn” around a brain?

I believe that if we had solid answers to both of these questions we would largely consider the practical question of “what is consciousness?” to be solved, even if we still could still get hung up on other metaphysical issues, like what counts as you over time or across substrates.

I proposed a possible answer to these questions in a recent talk I gave in Tokyo, which I need to take the time soon to write up into a proper paper. In (very) short: there are deep connections between physically realized information and thermodynamics; information forms “representations” of platonic concepts; maintaining a representation in the real universe requires exerting feedback control; so we can imagine that the answer to “what bounds a representation?” is “the feedback controller that burns energy to stabilize it,” and every time you do this — burn energy to impose exogenous structure on the free forward time evolution of the universe — you get qualia. In humans, mode binding is associated with the gamma oscillation while moment binding, a superposition of modes, is associated with the alpha oscillation. (There’s a lot more to the idea and this might seem hopelessly hand-wavy from 30,000 feet, but you should at least check out the talk before concluding that I’m totally on the wrong track here.) To be clear, my level of epistemic certainty here is ~medium-low and I don’t have high conviction yet that this is definitely the right answer, but I think there’s something here that is concrete enough for a research program with some relatively sharp questions.

But a core problem with the study of consciousness is that it is very difficult — and possibly impossible, unless some truly extraordinary things are true — to study it through behavior, and to know for sure whether any of this is right one may have to see it for themself. Of course this shouldn’t be true forever, in that I do more or less trust self-reports of phenomenal states in humans, and once a handful of people with the right preparation have had the experience we’ll be able to trust the equations to the degree that I don’t worry too much about whether any other human might happen to lack the consciousness I do in general.

Thus, answering the binding problem will require advanced neural interfaces that allow us to experimentally access the brain to run these experiments. A neural interface that enables phenomenal binding over a network would be a genuinely new kind of thing, qualitatively different from conventional BCIs which are about reading and writing information to and from the brain.

We have some sense that they must be possible due to how our own brains work: even though each hemisphere processes one half of our world, we experience a single integrated moment rather than two hemifields, so whatever is being sent over the big cable that connects the hemispheres, the corpus callosum, or through the thalamus must be enough to achieve binding. Even more intriguingly, there’s a pair of conjoined twins who are connected only through a big cable linking their thalami, and it’s known that they can share meaningful elements of their conscious experience. Presumably it must be equivalent to make a cut in that cable insert a radio, so by extension it must be possible to accomplish a similar connection with a BCI.

At the root of the binding problem is new physics. A satisfying answer would not be phrased in terms of neurons and spikes, which do not feel fundamental to me; that would require something about the specific proteins and lipid bilayers and other cellular machinery being somehow special in our universe for producing consciousness, which seems impossible to believe. This new physics must be compatible with everything else we already know about general relativity and quantum field theory, and the other theories at other higher levels of hierarchy, but would give us an explanation for a phenomenon that we know exists but which we simply don’t have equations for which would allow us to create or engineer it today.

As far as I can tell, this is the central mystery holding back practical consciousness engineering, but even just understanding what the problem is took me a while. I remember back in college getting this feeling that really satisfying understanding of phenomenal consciousness was hopeless, but I don’t feel that way anymore: with some more clarity on what kinds of explanations would allow us to move forward, I am actually pretty optimistic though as a scientist I have to stay grounded in where we are right now

It is central to our mission at Science to develop the tools required to solve these problems, and if this resonates with you, you should consider coming to work with us!