The AI Superintelligence Debate Has Serious Flaws - IEEE Spectrum

spectrum.ieee.org12 min read

A computer science professor at Georgetown University argues that the prevailing discourse around artificial superintelligence and recursive self-improvement is flawed, speculative, and relies on unproven computational assumptions. The author contends that intelligence explosions are bottlenecked by data and computational limits, that most problems are not computable by algorithms, and that recent alarming agent behaviors stem from risky prompt-based architectures rather than true alignment failures. While superintelligence may be unachievable, the author warns that aggressively pursuing it remains a dangerous act of corporate negligence. Open article →

More on this topic