lpetrich
Contributor
The other kind of AI is bottom-up AI. One starts with a model that has variable parameters, and one then runs some optimization algorithm to find the best-fit parameter values. Artificial neural networks or connectionism is a version of this approach.
Statistical natural-language translation uses this approach, and it's what Google Translate and some other recent autotranslators use. It must be noted that one needs a big corpus of translations to work from.
This approach may fail if there was not enough to work from, like with the word "cowed".
Statistical natural-language translation uses this approach, and it's what Google Translate and some other recent autotranslators use. It must be noted that one needs a big corpus of translations to work from.
This approach may fail if there was not enough to work from, like with the word "cowed".