Some users praised Gary Marcus's argument on hybrid LLMs with symbolic tools as a fair point about backend capabilities, while others dismissed it sarcastically as word salad or absurd calculator analogies.
Based on 6 visible X reactions from 43 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@GaryMarcus That distinction matters. A lot of what we call “model capability” now is really the capability of the model plus the tools around it. Where should we draw that line?
@GaryMarcus Amazing prediction Gary
@GaryMarcus And this from the evident CEO of an AI training company. Not good, Bob.
@GaryMarcus next they'll say calculators are good at math its just buttons and a screen
@GaryMarcus that’s such a fair point, people really do skip over the backend.
@zeratizerata @GaryMarcus No without tools they are really gpt 3.5
this is confused. it’s not LLMs that are good at math; it’s LLMs merged with (neuro)symbolic tools that are good at math. amazing that people feel free to clutch their pearls without understanding even the basic technical details.
Some users praised Gary Marcus's argument on hybrid LLMs with symbolic tools as a fair point about backend capabilities, while others dismissed it sarcastically as word salad or absurd calculator analogies.
Based on 6 visible X reactions from 43 accounts; directional sample.
Ask a question below.
Published answers will appear here.