If you think for a moment you will guess that "how fast is it getting smarter" probably depends on a bunch of shit, but also on "how smart is it right now" There's a relationship, in other words, between the function and its derivative. 4/
-
-
Replying to @amolitor99 @merrickdeville and
DiffEq is the study of how to take a such relationships and find solutions (either an actual formula, or a way to calculate efficiently) what the actual functions are. So a solution here is a way to calculate how smart our AI will be at arbitrary points in the future. 5/
1 reply 1 retweet 11 likes -
Replying to @amolitor99 @merrickdeville and
Fun fact: In the real world almost all solutions to this class of problem are periodic. Solutions typically start out with exponential growth, and then settle into a back-and-forth swinging. Or they drop to zero and stay there. 6/
1 reply 1 retweet 14 likes -
Replying to @amolitor99 @merrickdeville and
There are other possibilities, but none of them are "my AI gets infinitely smart" fin/
2 replies 2 retweets 17 likes -
Replying to @amolitor99 @merrickdeville and
The naive extrapolation of rates of change is so easy to dunk on It's like the xkcd joke about "Why I shouldn't be allowed to officiate weddings"
1 reply 3 retweets 16 likes -
Replying to @arthur_affect @amolitor99 and
"One minute ago you had zero husbands. Now you have one husband. So one minute from now you'll have two husbands Next year you'll have 525,600 husbands"
2 replies 1 retweet 21 likes -
Replying to @arthur_affect @amolitor99 and
Years ago, I saw an article saying that ten people had been eaten by alligators in Florida that year, compared to just one the year before. I noted that, at that rate of increase, it’d be less than a decade before everyone in the world had been eaten by alligators in Florida.
1 reply 2 retweets 17 likes -
This Tweet is unavailable.
-
The thing to note is that these *aren't* "intelligent people who have been successful in the tech realm", for the most part. Yudkowsky is a high-school dropout who has literally never released a working piece of code. Scott Alexander is a psychiatrist.
2 replies 1 retweet 26 likes -
This Tweet is unavailable.
They have a lot of fans among actual important people with important day jobs, that's why you've got all these CEOs signing the petition to support Scott
-
-
Replying to @arthur_affect @merrickdeville and
I must add, as far as I know, they are more influential among investors and marketing people than programmers, especially those who work on AI.
1 reply 0 retweets 6 likes -
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.