I guess some researchers will now call for "biologically diverse training datasets". Talk about missing the point.
This is really entertaining, even if you don't know much about chess (oh, hi! I was lost at Sicilian Najdorf Defense)
Anyway, the latest ChatGPT is scary good at some surprising tasks. And when that's not enough, it just fuggin' cheats 💀
Dear Fans and #Friends of others #bubbles,
As we may have #gathered, everyone's #experience, #priorities and view of #reality is VERY elephant ... eh wait ... #different. Despite this, we have a #commonality of #interchange, widely #accepted and #comprehensible.
Despite:
- #Different ages
- Many #languages
- Biases
- #Affiliations
- #Regional #Socialisation
AI (Awful Imbecility) is just a data sorting. Useful and also impractical for critical widened models of #context. Who knew that? Not #CheatGPT for Far Shore.
People have been trying to find meaning in the cracks of boiled turtle shells, the entrails of animals, the flight path of random birds, the lines of the hand, the cold trajectories of the stars crossing the night skies, and so on and so forth ad nauseam for like forever.
But trying to find meaning in the output of autocomplete is a new one.
I call this, "autocompletomancy".
And Lilian leads a team on AI Safety, shake my damn head.
Via @timnitGebru
In fairness, ChatGPT could replace Tom Friedman, Pamela Paul, David Brooks, Maureen Dowd, Maggie Haberman, and Peter Baker, and the bullshit level of the New York Times would not increase.
We should call it #CheatGPT.
It steals our works but gives no credit or compensation.