Exponential Knowledge Without Training

John Ball
Pat Inc
Published in
27 min readSep 3, 2019

--

Linguistic models scale exponentially when taught; NLP training data does not.

Linguistic models add exponential knowledge: that’s good. The data science training model, by comparison, is slow: that’s bad.

The “data” model promised effective NLP (natural language processing) given just “more data” and later, perhaps, AGI (artificial general intelligence). But data availability is terribly limited compared to the scale of a natural language and that possibly explains why the data model doesn’t scale to conversations.

I’ll use English to show the scale that machine learning systems need to deal with. For NLP to learn the meaning of words in a language, it needs to understand the relations, but today’s systems skip that step — in preference for annotated data, unannotated data, combinations or artificial representations with vectors from statistical sources.

Today’s analysis is long because languages are complicated and I want to establish some ballpark estimates of the scale of a human language to compare with the size of data. If I just said that in 3 days, I added approximately 9.16 x 10²⁹⁴⁴ unique and verifiable word sequences to my system, with the addition of only a few hundred associations, you may be sceptical. I would be, too, and so I’ll show the workings below.

--

--

John Ball
Pat Inc

I'm a cognitive scientist working on NLU (Natural Language Understanding) systems based on RRG (Role and Reference Grammar). A mouthful, I know!