this post was submitted on 31 Dec 2023
-2 points (45.0% liked)

Showerthoughts

29723 readers
2406 users here now

A "Showerthought" is a simple term used to describe the thoughts that pop into your head while you're doing everyday things like taking a shower, driving, or just daydreaming. A showerthought should offer a unique perspective on an ordinary part of life.

Rules

  1. All posts must be showerthoughts
  2. The entire showerthought must be in the title
  3. Avoid politics
    1. NEW RULE as of 5 Nov 2024, trying it out
    2. Political posts often end up being circle jerks (not offering unique perspective) or enflaming (too much work for mods).
    3. Try c/politicaldiscussion, volunteer as a mod here, or start your own community.
  4. Posts must be original/unique
  5. Adhere to Lemmy's Code of Conduct-----

founded 1 year ago
MODERATORS
vatlark@lemmy.world
forkball@lemmy.world
quinten@lemmy.world
Thekingoflorda@lemmy.world
-2
Sleep is when your brain's LLM retrains on the day's conversation (lemm.ee)
submitted 10 months ago by intensely_human@lemm.ee to c/showerthoughts@lemmy.world
1 comments fedilink hide all child comments
 

I’ve noticed ChatGPT gets less able to do precise reasoning or respond to instructions, the longer the conversation gets.

It felt exactly like working with a student who was getting tired and needed to rest.

Then I had above shower thought. Pretty cool right?

Every few months a new ChatGPT v4 is deployed. It’s got new training data, up through X date. They train up a new model on the new content in the world, including ChatGPT conversations from users who’ve opted into that (or didn’t opt out, can’t remember how it’s presented).

It’s like GPT is “sleeping”, to consolidate “the day’s” knowledge into long term memory. All the data in the current conversation is its short term memory. After handling a certain amount of complexity in one conversation, the coherence of responses breaks down, becomes more habitual and less responsive to nuance. It gets tired and can’t go much further.

you are viewing a single comment's thread
view the rest of the comments
[–] cheese_greater@lemmy.world 2 points 9 months ago* (last edited 9 months ago)

I wonder if it has more to do with the fact that the more inputs are available, the more complicated and prone to error something is. Also, ChatGPT doesn't "get tired" or at least not in the human meaning, I guess it could exhaust the memory of whatever its running on would be the closest analogy if my understanding is correct...

Re:complexity, this is true of all things, and computer/programming stuff is no different. The more complex it is or the larger the dataset, the greater potential for noise vs signal