Gregory's Server@Gargron Then we can agree that at this stage AI safety is another trick of marketing.
|
5 comments
 @dimillian @Gargron AI can be enormously problematic without it actually being intelligent  @dimillian You should maybe read up on how AI is being used for spam, misinformation, propaganda, impersonation, to name a few  |
@dimillian @Gargron sentience is orthogonal to intelligence. as loss functions continue to improve, new capabilities will spontaneously emerge corresponding to the improving predict-the-next-word ability, that make this externally observable improvement possible.
unless this process hits a wall at some point due to regulation, lack of progress in chips/algorithms, some kind of diminishing return, etc, eventually it will result in generalized intelligence superior to our own.
thanks for IceCubes by the way.
@dimillian @Gargron sentience is orthogonal to intelligence. as loss functions continue to improve, new capabilities will spontaneously emerge corresponding to the improving predict-the-next-word ability, that make this externally observable improvement possible.
unless this process hits a wall at some point due to regulation, lack of progress in chips/algorithms, some kind of diminishing return, etc, eventually it will result in generalized intelligence superior to our own.