Will future language models converge on "what Einstein would have thought of Many-Worlds?" before 2036?
➕
Plus
17
Ṁ535
2035
46%
chance

Get
Ṁ1,000
and
S1.00
Sort by:

The idea is: Imagine that best methods for integrating all text data were as well-understood as using odds ratios to aggregate bits of evidence. An LM properly trained just for predictive ability should do that in the limit.

Technically such convergent answers can remain uncertain, though this one shouldn't, for a reasonable context and manner in which he were asked.