At specific tasks, this has already happened. AI beats humans at chess (since 1997), Go (since 2016), protein folding (Nobel Prize 2024), and graduate-level ...
At specific tasks, this has already happened. AI beats humans at chess (since 1997), Go (since 2016), [protein folding](https://www.nobelprize.org/prizes/chemistry/2024/press-release/) (Nobel Prize 2024), and [graduate-level science questions](https://epoch.ai/benchmarks/gpqa-diamond) (94.1% vs. 65% for PhD experts). That part isn't a prediction — it's a fact. The real question is whether AI will achieve *general* intelligence that exceeds human capability across the board. On that, predictions range from "within five years" to "possibly never." Nobody knows — and confidence about timelines has been a poor predictor in either direction since the field began.
Verdict: Narrow superhuman AI is here. General superintelligence remains an open question. AI already surpasses humans at specific tasks — and the list is growing fast. Whether it achieves general intelligence that exceeds human capability across the board is genuinely unknown, with credible experts on every side. The people who say "five years away" and the people who say "impossible" are both making predictions they can't back up. The ones worth listening to are the ones who say "we don't know, and here's what we should do about that."