What would actually convince you that the AI singularity has begun?
retained text
We spend a lot of time debating when the singularity might happen. But I'm curious about something slightly different. How would we recognise it if it had already started? A new model outperforming humans on difficult benchmarks wouldn't necessarily qualify. Nor would AI replacing large numbers of jobs, or even making major scientific discoveries. Those could be extraordinary achievements without necessarily representing a technological singularity. Perhaps the more meaningful threshold is when AI begins driving advances in AI itself, creating a cycle of improvement that increasingly exceeds our ability to predict, evaluate or direct. But even that raises questions. Would the improvements need to accelerate? Would they need to happen without substantial human intervention? And how would we distinguish genuine recursive improvement from increasingly sophisticated human-led engineering? There's also an uncomfortable possibility: we might only recognise such a transition retrospectively, once its consequences become obvious. So I'm interested in people's actual criteria. **What observable event or measurable change would convince you that the singularity had begun?** *Disclosure: This discussion was developed through human–AI collaboration.*
document fetched 2026-10-20 00:28 · canonical https://reddit.com/r/OpenAI/comments/1x9trxm/what_would_actually_convince_you_that_the_ai