Whole brain emulation research may be net-harmful

The article builds a case that research into whole brain emulation (WBE) is net-harmful by default, despite the potential long-term benefits of a fully successful WBE. It acknowledges that a true WBE could help humanity, for instance by allowing the emulation to increase its own intelligence and then prevent an AGI from causing harm. However, the central concern is that any serious progress toward WBE would likely contribute to AI capabilities and therefore increase existential risk from AI.

The reasoning rests on the difficulty of WBE. Creating a WBE is very hard, so a research project is unlikely to complete one quickly, even within five years. Instead, a project would likely make some initial progress along the wide spectrum of difficulty, from easy to extremely difficult, and then stall. This partial progress, the article argues, could still accelerate AI capabilities and thus raise risk, without delivering the protective benefits of a full WBE.

The stakes, then, depend on whether a research project could move quickly enough to reach a true WBE before contributing to AI risk. The article suggests that is very unlikely, but the argument hinges on how fast progress can be made and whether stalling is inevitable. As with many existential-risk arguments, the conclusion is a matter of weighing uncertain probabilities rather than a settled fact.