Comment by throwaway48476
2 months ago
It would be helpful to tell users that it's just a model producing mathematically probable tokens but that would go against the AI marketing.
2 months ago
It would be helpful to tell users that it's just a model producing mathematically probable tokens but that would go against the AI marketing.
Telling people who are playing slot machines “it’s just a random number generator with fixed probabilities in a metal box” doesn’t usually work either
I feel like the average slot machine user is _far_ more aware of this than the average LLM user is of the nature of an LLM, tho. A lot of laypeople genuinely think that they think.
Also chatbots are explicitly designed to evoke anthropomorphizing them and to pull susceptible people into some kind of para-social relationship. Doesn't even have to be as obviously unhealthy as the "LLM psychosis" or "romantic roleplay" stuff.
I think the same thing is also relevant when people use chatbots to form opinions on unknown subjects, politics, or to seek personal life advice.
I've tried that, it doesn't work. They want to hear that from a famous person & all the famous people are telling them these things are going to take all of their jobs & then maybe also kill everyone.
And you’re a sack of meat and neurons producing learned chemical responses to external stimuli. Now tell me how useful that is.
https://www.mit.edu/people/dpolicar/writing/prose/text/think...
> Meat sounds. You know how when you slap or flap meat it makes a noise? They talk by flapping their meat at each other. They can even sing by squirting air through their meat.
That's good
I find it more useful to think of myself as part of a wave function splitting thousands of times a second.