Mike Janssen @mjanssen.bsky.social

I agree that LLMs don't "understand" anything but disagree that they're not good for analysis. Can be very useful in working with datasets, finding trends and outliers, etc.

3 replies 0 reposts 6 likes

C.V. Danes @cvdanes.bsky.social [ View ] Jun 25, 2024 at 17:16 UTC

Yes. Machine learning is good at processing large data sets and surfacing patterns, which a human would then investigate to uncover why. Getting rid of the human is taking the pattern at face value, with predictable results.

3 replies 0 reposts 34 likes

@flatline42.bsky.social [ View ] Jun 25, 2024 at 17:51 UTC

Are you using LLMs directly for that or are you also incorporating other machine learning or other solutions on clean usable data? We've had machine learning for a long time. I'm looking at moveworks and they're using LLMs to help drive traditional API calls in other platforms and that works well.

2 replies 0 reposts 2 likes

Andy Craig @andycraig.bsky.social [ View ] Jun 25, 2024 at 17:48 UTC

Granted, "analysis" was an imprecise word on my part. They can do some things you could call data analysis in one sense of the word. What they lack is abstract reasoning of any sort, including the abstract ideas of true or false or cause and effect. The lights are on but nobody's home.

0 replies 0 reposts 1 likes