Brilliant example of the pattern-matching limitations of ChatGPT (thanks to Tim Hazan): “What letter comes next O T T F F S S?” - first it gets the right answer, then justifies with the wrong explanation, then offers a wrong conclusion following its mistake @GaryMarcus

Dec 11, 2022 · 1:50 PM UTC

It sometimes seems impressive but moments like this, which there are “many”, help to explore its capabilities and limitations. Some may say that’s it for AGI, but surely this isn’t the AGI I am looking for.