2 ms·

There is probably a 256-byte program that simulates a universe in which a superhuman intelligence evolves. Of course the memory and runtime requirements of that

by bloak 6y ago

There is probably a 256-byte program that simulates a universe in which a superhuman intelligence evolves. Of course the memory and runtime requirements of that program would be prohibitive, and the superhuman intelligence would not know any human language, because there are more than 2048 bits of entropy in the basic vocabulary of any language (even Toki Pona, I would guess).

exikyut 6y ago
Hmmm. This makes me wonder if, in the same way we've been technically-erroneously been hyperfocusing on CPU speed and quantity of memory in system-measuring contests and benchmarks, we've developed a fixation for how little RAM or CPU we can use while doing something interesting. I don't see the same sorts of fights developing around how much disk space we use or what the specs of all the systems we used for preprocessing are. Demos are supposed to be about conquering limitations, right? Inspiring creativity and all that? Well, why can't we add those sorts of constraints in, then? Aren't we smart enough? :P We might even build consciousness one day if we're honest about how much total resources we're using. That way we would actually have to reason about all the bits (no more "GPUs are really *that complicated* that what you're describing is impossible") and there would be no more black boxes. Demos would be interesting.
ajuc 6y ago
> There is probably a 256-byte program that simulates a universe in which a superhuman intelligence evolves I can even write it :) x = 0 while (true): eval(int_to_code(x)) x += 1 of course the runtime will be probably quite long EDIT: I forgot about the stop problem. But that's easy - stop the eval when memory repeats. x = 0 prevStates = None while (true): state = x prevStates = set() while(not state in prevStates): prevStates += state state = eval_1_step(int_to_code(state)) x += 1 int_to_code and eval_1_step should be part of the language What really blows my mind is that this basically makes the multiverse (everything that could happen - happens in some other universe).

by bloak 6y ago

exikyut 6y ago

Hmmm. This makes me wonder if, in the same way we've been technically-erroneously been hyperfocusing on CPU speed and quantity of memory in system-measuring contests and benchmarks, we've developed a fixation for how little RAM or CPU we can use while doing something interesting. I don't see the same sorts of fights developing around how much disk space we use or what the specs of all the systems we used for preprocessing are. Demos are supposed to be about conquering limitations, right? Inspiring creativity and all that? Well, why can't we add those sorts of constraints in, then? Aren't we smart enough? :P We might even build consciousness one day if we're honest about how much total resources we're using. That way we would actually have to reason about all the bits (no more "GPUs are really *that complicated* that what you're describing is impossible") and there would be no more black boxes. Demos would be interesting.

ajuc 6y ago

> There is probably a 256-byte program that simulates a universe in which a superhuman intelligence evolves I can even write it :) x = 0 while (true): eval(int_to_code(x)) x += 1 of course the runtime will be probably quite long EDIT: I forgot about the stop problem. But that's easy - stop the eval when memory repeats. x = 0 prevStates = None while (true): state = x prevStates = set() while(not state in prevStates): prevStates += state state = eval_1_step(int_to_code(state)) x += 1 int_to_code and eval_1_step should be part of the language What really blows my mind is that this basically makes the multiverse (everything that could happen - happens in some other universe).