Finished watching the video on tokenization by @karpathy and, apparently it turns out that ChatGPT 3.5 is unable to read any continuous text wrapped within the special token symbols: <|texthere|> .
Putting white spaces between the text: <| text here |> makes it normal.
“This release is the culmination of many hours of collective effort to create a single file that compresses the visual information of humanity into a few gigabytes.”
Just for fun on lunch break, decided to see what would happen if I asked #DALLE to expand "Starry Night" with the caption "Oil on canvas painting of a small town in the south of france just before sunrise." (higher res version in reply). Pretty neat!
There's still this illusion that "AI" is somewhat intelligent: these are models fitted to large datasets using non-linear optimizations, hoping to replicate their distributions.
While these are incredibly useful tools, there is nothing we would consider human-like intelligence.
In 1935 physicist Erwin Schrödinger published his famous physics thought experiment, Schrödinger's cat, which presents a paradox in which a cat in a box is somehow simultaneously both alive and dead.
We’re thrilled to announce that the first results from Fermilab’s Muon g-2 experiment strengthen evidence of new physics! #gminus2
https://t.co/tUx4ojzIps