[LINK] Speed superintelligence?
From Toby Ord:
Tool assisted speedruns (TAS) are when people take a game and play it frame by frame, effectively providing super reflexes and forethought, where they can spend a day deciding what to do in the next 1/60th of a second if they wish. There are some very extreme examples of this, showing what can be done if you really play a game perfectly. For example, this video shows how to winSuper Mario Bros 3 in 11 minutes. It shows how different optimal play can be from normal play. In particular, on level 8-1, it gains 90 extra lives by a sequence of amazing jumps.
Other TAS runs get more involved and start exploiting subtle glitches in the game. For example, this page talks about speed running NetHack, using a lot of normal tricks, as well as luck manipulation (exploiting the RNG) and exploiting a dangling pointer bug to rewrite parts of memory.
Though there are limits to what AIs could do with sheer speed, it’s interesting that great performance can be achieved with speed alone, that this allows different strategies from usual ones, and that it allows the exploitation of otherwise unexploitable glitches and bugs in the setup.
I think this is a great metaphor of what a boxed AI would be able to do to the gatekeeper.
http://lesswrong.com/lw/qk/that_alien_message/
That one is a good read, but way too noisy, wordy and convoluted for the simple point that OP’s making.
Ironically, it’s what a programmer does to AI when stepping through it with a debugger.
I feel like a new term should be coined. “Philosopher’s AI”, sort of like philosopher’s stone or philosophical mercury.
I’d like to point out that those kinds of insane speed runs aren’t purely the product of playing at low speed, they’re also the product of being able to reset time and try again.
(or do you really think that if, on the final level, after months of ultra-slow play, if the player realizes that they’re going to miss a jump and die that they’ll start all over again)
Speed isn’t everything, if a chimp had 2 years to ponder every 2 minutes its still unlikely to be able to outsmart the zookeepers. more likely it would just start to lose track of why it was doing anything.
You can leave a genetic algorithm running for a million generations but you’ll probably not gain anything notable after a certain point because it will get caught in local minima. There’s plenty of humans who just end up in a loop if left with a problem for a long time.
Yes, but these speed runs are impressive even compared with humans who get to reset time.
Not so surprising. This assumes perfect information about the world as well as great speed.
Perfect information? As far as I know all tricks used in most TAS videos were discovered by players, most of the time by accident. Kinda like how scientists discover(ed) the laws of physics.
The point is that these speed runs presumably involve backtracking. They can rewind time and explore different paths until they find one they like.
So do regular playthroughs, though; it’s a video game. The first paragraph still remarks on “how different optimal play can be from normal play.”
Wait, I think the link is missing.
It came from an email.
I’m not sure speed alone, by itself, is a solution. If you speed down a game by, say, 70% there would probably be no difference than if you sped it down by 90%, since there’s a limit to what the character can do in a given second. Mario, for instance, once you jump, there’s not much to do until he actually lands.
Suppose the same would happen if we had the capability to speed down time in our actual lives. Sure you could dodge bullets and win F1 races from time to time, but the actual day-to-day tasks, that take the majority of our time wouldn’t be improved much. If you need to eat lunch, eating it in an optimal way won’t give you much advantage in comparison to regular people that don’t take the fork to their mouths following a perfect parabola.
Mario games let you change momentum while jumping, to compensate for the lack of fine control on your initial speed. This actually does matter a lot in speed runs. For example, Mario 64 speedruns rely heavily on a super fast backwards long jump that starts with switching directions in the air.
A speed run of real life wouldn’t start with you eating lunch really fast, it would start with you sprinting to a computer.
And if I needed to do research, finish writing a book, do computer based work for a company, etc… I could probably arrange things to do this much faster and make much more money.
Being able to eat while parkouring to your next destination and using a laptop at the same time might. And choosing optimally nutritious food. Even if you did eat with a fork, you wouldn’t bring the fork in a parabola, you’d jerk it a centimeter up to fling it towards the mouth, then bring it back down to do the same to the next bite while the previous is still in transit.
Not really relevant to real life, but there can be large differences.
Standardized IQ tests have specific times for completion for a reason. With more time (and the conscientiousness to actually use it) everyone would do better at them. So a speed increase is an IQ increase.
I think of my mental processing as a kind of information metabolism, with an equivalent of hunger: boredom. So I expect a smarter/faster thinker would need more input to process, much like a more muscular body needs more food. That, at the very least, should cause a smarter/faster thinker to interact with its environment substantially different from a standard rate one.
Link?
In an email. The links are just those included in the post.