AI safety paper discussion
This Thursday at 7pm in Culver City covering Anthropic's circuits research on mechanistic interpretability and neural network internals.
Pizza provided. Great mix of backgrounds all focused on understanding how to build safe AI systems.
@SyntheticBeef The moment this works
It will be better for the labs to do it themselves and no longer share the models with the public for less than the amount they can make themselves
@c_grutter@Aella_Girl@envisionedluna I there are people everywhere who would love to attend cnc orgies.
To make it actually happen you have to be extremely high openness, agency, and have an autistic obsession with sex.
There are few places outside of the rationality community this could have happened.
@c_grutter@Aella_Girl@envisionedluna The doomers are not in charge of the ai apocalypse, they want to shut down AI research.
There is a lot of tension between the people who work at the labs what are in charge of the apocalypse, and Aellas doomer friends who want them to stop.