Mind reading is no longer science fiction
Mind reading is no longer science fiction.
This is downstream of the metaphysical belief that the world is not seen directly by your brain. What you experience as the world is generated by your brain.
The human organism has sensors–eyes, ears, nerves–that are streaming information to the brain. The brain processes the inputs. For example, in the visual cortex, the information from the eyes is processed and generates objects that you experience.
Whether the brain itself generates consciousness invokes the hard problem of consciousness. This is not important. What we care about is that the contents of consciousness directly map onto the physical substrate of the brain, something we can observe. The existence of a mapping has been extensively shown in neuroscience through all kinds of brain imaging techniques. Libet’s experiments are an early example.
The brain is a world generator and we are caught in its simulation. Consider this thought experiment: you take psychedelics and every color becomes more intense.
Have the light waves reflected off of all objects suddenly changed? Obviously not. Your brain has been temporarily changed by a chemical substance and the world being generated by your brain has changed.
You experience a world downstream of your brain. The same brain that clearly generates your dream generates your waking experience, although the waking experience is bounded by some kind of external information (the universe). It might be more useful to think about the universe as a generator of information rather than a large empty space with things in it.
We have established that mental phenomena, including thought, are correlated with patterns in the brain. Once you see this, mind reading is simply a matter of mapping the brain’s activity onto natural language.
How do you read brain activity? This is what a brain-computer interface does. Noninvasive interfaces pick up on electrical signals emitted by the brain that correlate with brain activity. We therefore have data representing brain activity. The question is whether we can find a mapping that takes brain activity as input and expresses the corresponding thought in natural language as output.
How do you approach that problem? We need to discover a function that can generalize data to produce natural language. It needs to be able to reason about the world.
We have found that LLMs do exactly this. They take a corpus of data consisting of words (or tokens) from the internet, books, other media. They then employ a transformer architecture, which is simply one form of a computation that can model a wide range of functions (including those in natural language). All it took was enough data and compute.
Now, suppose you gave a transformer-based model a gigantic amount of brain-activity data. Orders of magnitude more than you might initially imagine, perhaps an internet-scale quantity. Could it reason about the space of brain activity and transform that activity into thought? If so, we would have a real-time mind reading machine.
That is the read operation. I haven’t yet discussed the write operation: What if you could write thoughts to the brain?
Generating thoughts in your first-person experience of the world is still thinking too small. If I can change anything about the contents of experience, I can change perceived physics or time. I can make someone experience several localized versions of consciousness at once. Once you unlock the ability to write to the brain, many kinds of distortions and alterations of experienced reality become possible.
This write capability has not yet been solved. It is early and right now requires surgery. But we will soon have non-invasive technologies that can take our commands about reality and convert them into experience. We will be our own god.
The non-invasive write operation remains an open question.
There are two ways of trying to understand the world you inhabit.
One is going out there. Space travel, for example. Do you learn about the world by going outward? Do you visit the planets and examine the galaxies? Do you discover new molecules, new species, new physics?
The second is going in here *points to brain*. What if you understood the world from inside out? What if you examined the thing that experiences the generated world? Is the brain the more interesting place to start?
In the nondualist tradition from which I emerge, the answer is that you learn more from going to the inside.
It should be clear that interfacing with the brain is the greatest technological project of our time. It is also the greatest philosophical project we have ever embarked on.