Technology

70995 readers

3717 users here now

This is a most excellent place for technology news and articles.

Our Rules

Follow the lemmy.world rules.
Only tech related news or articles.
Be excellent to each other!
Mod approved content bots can post up to 10 articles per day.
Threads asking for personal tech support may be deleted.
Politics threads may be removed.
No memes allowed as posts, OK to post as comments.
Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
Check for duplicates before posting, duplicates may be removed
Accounts 7 days and younger will have their posts automatically removed.

Approved Bots

founded 2 years ago

MODERATORS

[email protected]

274

I'm looking for an article showing that LLMs don't know how they work internally (feddit.it)

submitted 6 days ago* (last edited 6 days ago) by [email protected] to c/[email protected]

78 comments fedilink hide all child comments

I found the aeticle in a post on the fediverse, and I can't find it anymore.

The reaserchers asked a simple mathematical question to an LLM ( like 7+4) and then could see how internally it worked by finding similar paths, but nothing like performing mathematical reasoning, even if the final answer was correct.

Then they asked the LLM to explain how it found the result, what was it's internal reasoning. The answer was detailed step by step mathematical logic, like a human explaining how to perform an addition.

This showed 2 things:

LLM don't "know" how they work
the second answer was a rephrasing of original text used for training that explain how math works, so LLM just used that as an explanation

I think it was a very interesting an meaningful analysis

Can anyone help me find this?

EDIT: thanks to @theunknownmuncher @lemmy.world https://www.anthropic.com/research/tracing-thoughts-language-model its this one

EDIT2: I'm aware LLM dont "know" anything and don't reason, and it's exactly why I wanted to find the article. Some more details here: https://feddit.it/post/18191686/13815095

you are viewing a single comment's thread
view the rest of the comments

[–] [email protected] 0 points 3 days ago* (last edited 2 days ago) (1 children)

There is a distinction between data and an action you perform on data (matrix maths, codec algorithm, etc.). It’s literally completely different.

Incorrect. You might want to take an information theory class before speaking on subjects like this.

I literally cannot be wrong that LLMs cannot think or reason, there’s no room for debate, it’s settled long ago.

Lmao yup totally, it's not like this type of research currently gets huge funding at universities and institutions or anything like that 😂 it's a dead research field because it's already "settled". (You're wrong 🤭)

LLMs are just tools not sentient or verging on sentient

Correct. No one claimed they are "sentient" (you actually mean "sapient", not "sentient", but it's fine because people commonly mix these terms up. Sentience is about the physical senses. If you can respond to stimuli from your environment, you're sentient, if you can "I think, therefore I am", you're sapient). And no, LLMs are not sapient either, and sapience has nothing to do with neural networks' ability to mathematically reason or use logic, you're just moving the goalpost. But at least you moved it far enough to be actually correct?

[–] [email protected] 1 points 6 hours ago (1 children)

It’s wild, we’re just completely talking past each other at this point! I don’t think I’ve ever gotten to a point where I’m like “it’s blue” and someone’s like “it’s gold” so clearly. And like I know enough to know what I’m talking about and that I’m not wrong (unis are not getting tons of grants to see “if AI can think”, no one but fart sniffing AI bros would fund that (see OP’s requested source is from an AI company about their own model), research funding goes towards making useful things not if ChatGPT is really going through it like the rest of us), but you are very confident in yourself as well. Your mention of information theory leads me to believe you’ve got a degree in the computer science field. The basis of machine learning is not in computer science but in stats (math). So I won’t change my understanding based on your claims since I don’t think you deeply know the basis just the application. The focus on using the “right words” as a gotchya bolsters that vibe. I know you won’t change your thoughts based on my input, so we’re at the age-old internet stalemate! Anyway, just wanted you to know why I decided not to entertain what you’ve been saying - I’m sure I’m in the same boat from your perspective ;)

[–] [email protected] 1 points 3 hours ago

loses the argument "we’re at the age-old internet stalemate!" LMAO