Re:Sam and Greg's response to OpenAI Safety researcher claims
by tbrownaw on Saturday May 18, 2024 @11:11PM
(#40403933)
It's essentially Pascal's wager.
[ Parent
| on HN ]
Re:Sam and Greg's response to OpenAI Safety researcher claims
by N0b8ez on Saturday May 18, 2024 @11:20PM
(#40403972)
What probability for extinction do you consider to be low enough that it effectively becomes a Pascalian wager? Pascal was also writing about a single person's fate (AFAICT), whereas this is about the fate of everyone.
Re:Sam and Greg's response to OpenAI Safety researcher claims
by kevindamm on Saturday May 18, 2024 @11:40PM
(#40404090)
Not parent, but perhaps it was an analogous interpretation: believe in AGI and AGI is possible vs believing in AGI and AGI is not possible, likewise for not believing.
Perhaps some people think believing in the possibility of AGI is justified because they are positioning themselves to benefit from it (with no real loss if it doesn't pan out) vs acting like it's not possible then missing out if it is.
Meanwhile, others of us that recognize its limitations see their enthusiasm as just wanting to believe so badly.
Re:Sam and Greg's response to OpenAI Safety researcher claims
by null0pointer on Sunday May 19, 2024 @12:53AM
(#40404364)
The point is that the probability doesn’t matter because the outcome, total human extinction, is given infinite weight.
Re:Sam and Greg's response to OpenAI Safety researcher claims
by N0b8ez on Sunday May 19, 2024 @02:23AM
(#40404661)
Maybe Pascal argued that way for the human soul, but I don't think the AI risk argument needs infinite value to be at stake. Would you say that preparing for the risk of a solar flare or an asteroid impact is also a Pascalian wager?