LessWrong's Vals uses red-green swap to explain consciousness

In a viral LessWrong post, tutor Vals tackles the hard problem of consciousness using a simple thought experiment: ask someone to imagine what it would be like if their experience of red and green were swapped, while the outside world and their sensory inputs remain unchanged. The exercise is designed to point to qualia—the raw, subjective quality of experience that isn’t captured by describing light wavelengths or neural activity. Vals pushes a skeptical denialist to accept that such a swap changes something internal (experience) without altering external physical inputs or brain wiring. The denialist eventually agrees but then questions whether a simple image classifier could be considered conscious under the same logic, revealing deeper issues about what qualifies for phenomenal experience.

The conversation bridges philosophy and AI: if consciousness is more than processing signals, how do we distinguish between a human brain and an advanced AI? The denialist’s challenge—'You could do this to an image classifier'—forces clarity on what unique property humans have. Vals sticks to defining what people mean by 'consciousness' rather than proving its existence, making this a primer for those who deny having conscious experience. For AI researchers, the post highlights why the hard problem remains unsolved: we can simulate color swapping in a neural network, but does the network feel anything? This distinction is critical for debates on machine consciousness and alignment.