This short, 8-minute talk was delivered at the Philosophy of AI conference at the Technion (in Haifa, Israel) in June 2026, as part of a panel on consciousness and AI. In the talk I introduce a challenge to computational theories of consciousness, which is the standard account underlying claims of AI consciousness. Drawing on the concept of vagueness, I motivate the view that concepts for computational states are inescapably vague, whereas the concept of consciousness is sharp. I then try to show how it follows from this that the nature of consciousness cannot be computational. The talk is based on research I published a number of years ago (see e.g. here and here). More information can be found in the video’s description on YouTube.