4 ms·
I'm not sure that anyone's actually tested what happens when a lot of complex rule sets are allowed to overlap and each have access to virtually all information
by noduerme 15y ago
I'm not sure that anyone's actually tested what happens when a lot of complex rule sets are allowed to overlap and each have access to virtually all information in the world; but I'm not sure the result wouldn't be a form of sentience. Sentience can't really be measured other than by I/O... the whole point of a Turing test is to figure out what's between the ears and the mouth of something that looks and acts like a sentient being.
I'm not saying it's there yet; I'm asking when it will be, and how it will act when it is. If you took something with a brain the size of a lobster's, say a couple million neurons or a few thousand programs competing for processing cycles, stuck a billion eyes and ears on it and fed it terabytes of information for ten to fifteen years (and gave it a perfect memory plus access to a few hundred million email passwords), are you really sure its behavior would be predictable? We're not talking about a bunch of rails scripts here. This code is made to improve large portions of itself to begin with.
- icebraining 15y agoI'm not sure that anyone's actually tested what happens when a lot of complex rule sets are allowed to overlap and each have access to virtually all information in the world I'm pretty sure that they don't all have access to all information; if anything because that would make it much harder to scale. They probably have access to internal services/APIs which give them what they need, more or less like Google App Engine works. If even their browser, which is designed to run on a single machine by a single user, runs with a bunch of different processes which are limited by simple, well defined interfaces to each other and blocked from other IO, I'd say it's unlikely that they simply let their internal programs run 'freely'. If you took something with a brain the size of a lobster's, say a couple million neurons or a few thousand programs competing for processing cycles, stuck a billion eyes and ears on it and fed it terabytes of information for ten to fifteen years (and gave it a perfect memory plus access to a few hundred million email passwords), are you really sure its behavior would be predictable? Almost all software has unpredictable behavior, but to consider that a combination of email servers, web page crawlers, phone routers, speech recognizers and such will develop sentience just because they're running on the same network is frankly ridiculous. Yes, there are probably errors, strange behaviors that result from complex interactions, but to the likelihood of intelligence emerging from that is, in my opinion, rounded to zero. It's like locking a few thousand animals of different species in a room and expecting them to turn themselves into a single thinking entity, when they're probably just going to eat each other. This code is made to improve large portions of itself to begin with. It is? I've never read that.