I'm wondering if I'm right in thinking Shannon's noisy channel encoding theorem makes AI boxing even theoretically impossible.
Does that make sense? cc: @ESYudkowsky @gwern @FrameOfStack @VesselOfSpirit
(Based on a Facebook conversation with @RokoMijicUK (?) and @romanyam)
-
Show this thread
-
To clarify, the noisy channel encoding theorem should mean there is a theoretical prohibition on *completely* eliminating side-channels for an AI reading outside information, even if you can mitigate the channels arbitrarily well. (Even a Faraday cage only reduces the bandwidth!)
1 reply 0 retweets 1 likeShow this thread
Replying to @davidmanheim @ESYudkowsky and
It might be made cryptographically impossible; you're confidant up to cryptographic assumptions that it won't be able to use side-channels to optimize, due to how they're scrambled.
11:47 AM - 9 Feb 2018
0 replies
0 retweets
1 like
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.