Part 9/11:
While recognizing the incremental nature of AI development, Shapiro notes that progress toward superintelligence will likely happen gradually rather than through sudden leaps. Each iteration improves capabilities, yet a catastrophic leap to superintelligence remains distant, users must remain cautiously optimistic but realistically skeptical about the immediate threat.
He briefly praises recent efforts by OpenAI, such as constitutional AI—a step toward more aligned autonomous systems—while criticizing their continued focus on reinforcement learning with human feedback as insufficient.