In 2002, Mitch Kapor bet $20k that no AI would pass the Turing Test before 2029 (longbets.org)

In 2002 Mitch Kapor famously wagered $20,000 that no machine would pass a properly administered Turing Test before 2029. The bet centers on Turing’s original, text-only formulation: a judge interacts with a human and a computer exclusively via instant messaging, probing everything from facts to personal history and subjective experience. Kapor (and critics he cites) argue that while machines have achieved striking narrow successes (Deep Blue, Jeopardy, high SAT-like scores, domain-specific expert systems), these rely on explicit, codified knowledge and brittle heuristics—not the tacit, embodied, multi-sensory, emotionally rooted know-how humans use to weave coherent, imaginative life stories. An AI that truly impersonates a human would need to invent plausible lived histories and demonstrate flexible creativity across domains, not merely regurgitate scanned books or facts. Technically, the critique targets paradigms like Kurzweil’s scan-and-emulate brain strategy: reverse-engineering “massively parallel digital-controlled analog algorithms” or harvesting text corpora may miss hormonally modulated cognition, embodied perception, and tacit knowledge that are not written down. The implication for AI/ML is twofold: benchmarks that focus only on text or narrow tasks risk overestimating human-level progress, and genuinely humanlike performance may demand new approaches—embodied learning, richer multimodal grounding, or architectures that capture tacit, experiential structure—rather than scale‑up of current language models alone.