-
The suggestive names error.
-
Mistaking the structure of one's language for the structure of the world.
1Something on the lighter side: the Mastermind box design.
Think less “family night,” and more “I will find the hidden structure of your choices, and then I will destroy you.”
No questions about LLM usage and attitudes?
Some questions off the top of my head I'd like to see the results of (no need for this to be in the standard LW poll):
-
How people are using LLMs.
-
Which ones they use.
-
How useful they find them.
-
How much they pay for them.
-
How seriously they take LLM welfare, consciousness, moral standing, etc.
-
Whether an LLM has ever talked them into a delusion (or a belief that everyone they tell it to says is a delusion), or had some other dysfunctional effect.
-
...
Our incredibly badly designed brains have some subprocesses that take a bit of predictive machinery lying around and repurpose it to send a control signal
I like this, and will show it to some of my colleagues who are also sceptical of the FEP/ActInf paradigm.
Another possible interpretation is a fund-raising drive of the form "I would publicly commit to paying up to this much conditional on enough other people publicly committing to paying at least as much."
Another is "if LW (with its content as-is) had been paywalled from the beginning, what annual fee would you be happy to have been paying?"
Was this inspired by active inference?
I wondered the same thing. I'm not a fan of the idea that we do not act, merely predict what our actions will be and then observe the act happening of itself while our minds float epiphenomenally above, and I would be disappointed to discover that the meme has found a place for itself in Eliezer's mind.
If the office-worker fears that they will be fired if
The office worker has gone wrong already by panicking and failing to ask, "Is this actually from the boss?"
Always be asking "What am I looking at?"
Claude said to keltan:
This is a criticism I've seen circulating in AI safety discussions
Surely Claude is bullshitting? I haven't heard of Claude or any other AI going about reading the net in its spare time, the better to hold up its end of an informed and up to date conversation. Maybe that's something that AIs will be doing soon, but AFAIK not yet. However, it is the sort of thing that a knowledgeable human with their finger on the pulse of events would say, so Claude says this sort of thing.
1Yesterday I received this email:
————
Driver & Vehicle Licensing Agency
Dear Vehicle Owner,
Our system has identified that your vehicle tax will be due on Saturday, 06 December, 2025.
To ensure that your vehicle remains legally paid and complies with applicable traffic regulations, you are required to complete the renewal process before the due date
[Update Now]
Thank you for keeping your vehicle tax up to date and for helping us maintain safe and legal roads across the UK.
————
I own a car, and the DVLA and vehicle tax are real things, so my immediate reaction to this was "Is this a genuine message?" It wasn't, of course. But the scam email got me wondering: are the sort of people who fall for such things also the sort who think that the LLMs they talk to are people?
Can I ask you to unwind the fundamentals a step further, and say why you and neuroscientists in general believe the brain operates by RL and has a reward function? And how far down the scale of life these have been found?