Consensus / Non-Consensus

A judgment most people already agree with is “consensus”; one only a few hold is “non-consensus.”

This is common language in investing and startup circles. Consensus refers to a judgment most people in an industry already accept, such as “data is the bottleneck for embodied AI”; non-consensus refers to a judgment only a minority believes, not yet widely accepted. Investors often say that only a “non-consensus and correct” view can produce outsized returns, because consensus views are already reflected in prices and valuations. Many debates in embodied AI can be described with this pair of terms: humanoid versus non-humanoid, real-robot data versus simulation data, end-to-end versus layered architectures. A given judgment tends to shift from non-consensus to consensus as evidence accumulates, at which point anyone entering later generally can't capture the early-mover returns anymore. When reading financing news and interviews, this term is often used to explain why a founder chose a particular technical route.

ExampleA few years ago, “control multiple types of robots with one large model” was non-consensus; after models such as π0 and GR00T were released, it gradually became industry consensus.

Related

Technical Route Debate · Form-Factor Debate · Real-Robot-Data Camp vs. Sim-Data Camp · Embodied AI Bubble · Funding Rounds & Valuation

Sources

Contrarian investing - Wikipedia