A casual Reddit post from early June has pulled AI researchers into discussion after one user described predicting ChatGPT-style replies in their head during daily moments like deciding whether to let a cat out, all without opening the app.
“What would I do if I had 10x more agents?”
sure this sounds silly, but "what would someone wiser than me do here" or "WWJD" are extremely overpowered in their own ways, so I basically expect this to work
Industry voices note parallels to running several internal models at once, similar to asking what different personas might suggest, though no accuracy checks or user studies back the original claim.
No data yet shows how many heavy users experience this or how it connects to established ideas like predictive processing, leaving the anecdote as viral conversation rather than benchmarked finding.
Positive users appreciate insights on internalizing AI mindsets for thinking, while negative users mock the Redditor's claim of mentally simulating ChatGPT responses as absurd or unbelievable.
@shortmagsmle Going to start pitching to employers that I have an LLM in my head whose token usage comes free with my employment.
Accidentally re-inventing “thinking” from first principles
JUST IN: Redditor claims he can now “use ChatGPT” in his head & accurately predict what it would say.
@shortmagsmle This is funny but it’s really important to note that what he is doing is altering the way he thinks based on the outputs he constantly exposes himself to. More and more morons are going to do this. They are “ChatGPT”-ifying themselves by choice.
@shortmagsmle I don't believe redditors can ever gain consciousness. God would not permit such a cruel joke.
@shortmagsmle Lmao, incredible stuff.
@shortmagsmle It's reddit.
No one there thinks.
@shortmagsmle
@shortmagsmle thomas running type shit
No different than “imagine what I would do if I had 10x agency” then doing it
Good theory of mind exercise. But also parasocial AI syndrome
People build a sounding board of friends with varying opinions
Over time you can predict what your friends are going to say when you tell them things
@shortmagsmle No there’s actually something interesting here. They’ve internalized the AI’s frame of mind, in the same way you would a parent or friend. There’s n order effects from a large part of society internalizing same frame of mind. Accelerating the hive mind.
@shortmagsmle This will lead to people who, by their own self-regression, will more and more begin to truly believe that these LLMs are the height of thought not because they are but because these cattle have mirrored their thoughts completely to them.
@shortmagsmle @TLEPaz The first redditor to unlock internal narrative/dialogue and the best they could compare it to is AI. Amazing self report.
@shortmagsmle people just next-word predictors confirmed, ted chiang on unalive watch
@shortmagsmle @EudaimoniaEsq I think this might be the opposite of first principles
@shortmagsmle This sounds like a joke but what if this is the end result of LLMs... people relearning how to think for themselves? It would be the most ironic outcome and imo irony always wins...
@badgraphix @shortmagsmle Nah man. Gotta charge extra for that compute!
@shortmagsmle @Babygravy9 Artificial intelligence was running on carbon circuits long before anyone put it on silicon.
@whittomd @shortmagsmle I’d quibble with this statement as it assigns agency to the model. What I am suggesting is that people are “hacking” themselves to further confirmation bias and wishful thinking about the “power” and “genius” of the models.
@shortmagsmle From last principles
@shortmagsmle @DrunkRepub Does he ever run out of tokens?
A casual Reddit post from early June has pulled AI researchers into discussion after one user described predicting ChatGPT-style replies in their head during daily moments like deciding whether to let a cat out, all without opening the app.
“What would I do if I had 10x more agents?”
sure this sounds silly, but "what would someone wiser than me do here" or "WWJD" are extremely overpowered in their own ways, so I basically expect this to work
