Post #1895923

2023-05-28 17:25 UTC

LLMs are modeling outputs, not replicating a process, as a result the output looks the same but it isn&#39;t made of the same stuff. It is a plastic banana. There is nothing inherently wrong with a plastic banana but as soon as you claim you can use it to solve world hunger people are going to be upset and it doesn&#39;t matter how much it looks right.

Replies (7)

@amsomniac@mastodon.mit.edu 2023-05-28 17:29

@Vrimj@mastodon.sandwich.net https://www.youtube.com/watch?v=gVvhxwtFsnE #nowplaying

Open ##2783161
@zippy1981@hachyderm.io 2023-05-28 17:50

@Vrimj@mastodon.sandwich.net lots of people in the technical-professional class do in fact just copy and paste a something from Google and don't know what it's actually doing for some subset of their job. They effectively are doing what chat gpt does.

Open ##2783162
@strwbrryJen@mastodon.social 2023-05-28 20:04

@Vrimj@mastodon.sandwich.net very much this

Open ##2783167
@LWFlouisa@cyberplace.social 2023-05-28 23:40

@Vrimj@mastodon.sandwich.net Excellent, I needed that image to brighten my mood. Lets talk to a plastic banana.

Open ##2783168
@PavelASamsonov@mastodon.social 2023-05-29 03:38

@Vrimj@mastodon.sandwich.net damn I really like "modeling outputs vs replicating a process" as a way to articulate the distinction

Open ##2783169
@kevinrns@mstdn.social 2023-05-29 04:46

@Vrimj@mastodon.sandwich.net It is "word complete while typing" with entire stolen books as the next letter.

Open ##2783170
@Techronic9876@sigmoid.social 2023-05-29 14:37

@Vrimj@mastodon.sandwich.net this is why chain of thought prompting completely surprised researchers working on LLMs. It seemed like using chain of thought replicated a process, and it possibly does for very simple processes Figuring out how to do bigger processes is the big question now

Open ##2783171