3 Comments

A small side-note here: As you are thinking about getting back into tinkering with the models and experimenting with them again (I really liked your "let's explore the boundaries of what we can do with GPT-3" stuff back then) - this would pose an excellent opportunity to engage in the reverse to your proposal: it should be also a goal of the the proponents of the idea that you actually __can't__ do anything (really) harmful with current (open) models to actually attempt to disprove themselves. For example, take one of the uncensored (smaller) models like the dolphin series and test out to what outcomes they can actually contribute in order to gauge the size of danger potential more accurately. Just a thought on this ...

Expand full comment

This kind of narrative is so good, not gonna lie lol. Writing a fictional story that represents the conversations we see almost daily now is really nice. My writer brain was really engaged with this one. Well done, my friend!! After all, fiction is probably humanity's greatest invention and strength. Keep going, my friend!

Expand full comment

“Fear is the worst counsel” .. though advice to take when dealing with extinction level events. But it holds up.

Expand full comment