Show newer
niplav boosted

Prediction this makes: conflict theorists more often profess Knightian uncertainty about a thing

Show thread

Knightian uncertainty is applicable if there is a process explicitely optimising against your belief state

It is unfortunate that positive/negative can mean "presence"/"absence" and "being for"/"being against".

It is similarly unfortunate that "phobic" can mean both "afraid of" and "being against".

niplav boosted
niplav boosted

@vriska Number of pillows on your bed

niplav boosted

@vriska Number of blankets on your bed

how many metaculus questions resolved on technicalities?

niplav boosted

AI, AIPOL 

Even as dumb as they are, current LLMs do have some extremely limited emergent general intelligence capabilities.

The comparison to parrots is apt, because parrots are (generally) intelligent too. Don't knock parrots just because they aren't humans (and the parrots are currently smarter than our best AIs).

We also (as far as I'm aware) haven't found the limits of current approaches. We really don't know what's going to happen.

Show thread

In general, how good are humans at solving problems compared to the theoretical optimum?

hint: he's been working on……another project, let's say

Show thread

you know why gwern has been posting so little in the past ~4 years, right?

Why was 6 afraid of 7? Because 7 8 9 28 17639 576189438…

Sudden takeoff view: AI systems are like normal programming (where the system suddenly works ~much much better by fixing step 14/15)

Slow take off view: AI systems are like physical with graceful decline

tasteless joke 

Salamon deez nuts

Show older
Mastodon

a Schelling point for those who seek one