Joe Vinegar on Nostr: nprofile1q…zvc7v actually the current brand of AI, namely LLM, are much worse than ...

nprofile1qy2hwumn8ghj7un9d3shjtnddaehgu3wwp6kyqpqcgccspusk920mw2jrhyg4jj7qx9c6mjur2ugg73kd3s76c454g2sgzvc7v (nprofile…vc7v) actually the current brand of AI, namely LLM, are much worse than imperative programming: with programming the designer and coder know what's happening, and build the software accordingly. For LLMs (an application of Deep Learning) the "programming" is done statistically, by example, and trial-and-error, with almost no understanding at all. and the overall result, the "model", is completely opaque in its functioning: not even the creators can predict what will be the output.
Orders of magnitude less reliable and trustworthy of any software ever created before.