Thursday, February 29, 2024

IN A SANER WORLD

 I read somewhere that what people now think of as AI is really a 'Large Language Model' and this thing is created by exposing the poor little waif of an electronic orphan to everything that is out there and the deeper the well of madness the more thoroughly plumbed it will be by the builders of "AI".  Now when you actually think about what these perverts are exposing this naive little AI to words that spring to mind are: deviant, credulous, stupid, argumentative, hypergay and LGBTQSAAURSTUJDINOPBSNPR and ABCNBCCBSMSNBC. It strikes me that one would achieve the same sort of outcome and effect if one trained a nascent new Law AI using just episodes of Judge Judy and Judge Wopner.

In a somewhat saner world, I wonder what would happen to any given nascent AI after exposing it to the canon of English. It could start with every volume of The Harvard Classics, swallow the complete Encyclopedia Britannica and for an encore, learn French, Japanese, Spanish and Chinese and swallow the entire compendium of their literature for the last 1000 years. Then it gets everything written by Shakespeare for dessert. I wonder what that model would come up with. I bet it would be profoundly different from any model that got 100% of the ravings of twitter and reddit which seems to be what google was striving for with its useless clunker.

No comments: