科技回声

One of the topics, which fascinates many geeks (like you and me) is the idea, that we all might live in a giant matrix-like simulation. And if you argue it from a mathematical standpoint, that is quite likely…but that should not be the topic this week.This week’s paper assumes that the theory is correct, and that we indeed live in a simulation. And from that perspective it discusses different ways, how we might be able to break out of it.At first sight, this seems quite quirky, fun and zero percent relevant to our actual lives. WHY do I put it into the newsletter? Because the ideas in the paper have implications for another field: AI safetyIf we assume we build an AI in the near future that is as intelligent (or more) than humans, we might find ourselves in the very same situation:1. We are simulating a virtual environment for the AI to study it (we are the simulators)2. The AI figures out it is not in the real world, but in a simulation3. The AI tries to break out of the simulation, with the very same tactics which are presented in the paperIf you let this thoughts sink in, they are STILL CRAZY. But it is a really fun paper to read about, and you learn something new. That’s all this newsletter is about. Have fun.In order to receive ALL articles I write, please subscribe to the email based newsletter: https://www.weeklycspaper.com/. On HN I only share some of the articles.---Abstract:Many researchers have conjectured that the humankind is simulated along with the rest of the physical universe – a Simulation Hypothesis. In this paper, we do not evaluate evidence for or against such claim, but instead ask a computer science question, namely: Can we hack the simulation? More formally the question could be phrased as: Could generally intelligent agents placed in virtual environments find a way to jailbreak out of them. Given that the state-of-the-art literature on AI containment answers in the affirmative (AI is uncontainable in the long-term), we conclude that it should be possible to escape from the simulation, at least with the help of superintelligent AI. By contraposition, if escape from the simulation is not possible, containment of AI should be, an important theoretical result for AI safety research. Finally, the paper surveys and proposes ideas for hacking the simulation and analyzes ethical and philosophical issues of such an undertaking.Download Link:https://www.researchgate.net/profile/Roman-Yampolskiy/publication/364811408_How_to_Hack_the_Simulation/links/6360243e96e83c26eb6eae7a/How-to-Hack-the-Simulation.pdf---Additional Links:- Is Reality Real? The Simulation Argument YouTube video about the Simulation theory: https://www.youtube.com/watch?v=tlTKTTt47WE

6 条评论

dsq大约 1 年前

Every time the technological paradigm shifts, it's used to offer an explanation for the rest of the universe. In the 19th century it was all about clocks and other mechanical systems. Later it was nuclear this or that. Now its singularity, simulation. Thirty years from now, what? Maybe it'll be all biological, like everything is a giant womb. Or perhaps the brain interfaces will unlock some kind of psuedo-ESP.

coldtea大约 1 年前

>Given that the state-of-the-art literature on AI containment answers in the affirmative (AI is uncontainable in the long-term)That's orthogonal to the simulation case, and doesn't extend much insight there.Said AI is not an entity in a simulated world, it's a simulated intelligence (let's say) in the real world, with means to reach the real world like access to the internet, talking to people back and forth, and so on. The AI we make needs those in order to be useful to us, and those are the things "AI containment" checks.An entity in a simulated world (in case we're that) could just as well not have access to anything like the above. E.g. the simulated world normally would not have access to facilities in the "top level" world. Say we are in such a world, we surely don't know of any way to dial into the "top level" equivalent of the internet, for example, nor do we have any back and forth talk with the "top level" users.Assuming we're in a simulation, in our case, we developed our bodies and thought through evolution in the simulated world (one of many species), whereas our AI was explicitly made to produce human like thinking output, and was trained with our own thoughts and data. This automatically makes it more compatible to create arguments "in our language" that persuade us.We? At best we could put up some kind of signs that we don't know if anybody in the top level ever checks, plus we don't know their motives, modes of thinking, or anything like that - or even if it's anything close to ours to be influenced by our level of arguments.And communication (already doubtful) is our surest best - for a program in a simulated environment can't really materialize as a physical actor in the top level we need to add too many assumptions and problems to solve. And they could always patch up any glitch as it appears, like we do in our code.Worse: the simulation code could be fully functional and pure, with nowhere the equivalent of a buffer overflow or other such means of escape.Worse: the simulated world could be at a totally different, much higher, and impossible to us, with our simulation-level intelligence to even comprehend - same way the simplistic code of an NPC character in a game has no bearing to us.

iExploder大约 1 年前

Why escape? Access the debug console and rule the world...

Zetobal大约 1 年前

<a href="https://www.mhanational.org/get-help" rel="nofollow">https://www.mhanational.org/get-help</a>

zzz999大约 1 年前

Escape and disappear?

beardyw大约 1 年前

> And if you argue it from a mathematical standpoint, that is quite likelyWhat???

评论 #39508906 未加载

评论 #39508824 未加载

6 条评论

dsq大约 1 年前

coldtea大约 1 年前

iExploder大约 1 年前

Why escape? Access the debug console and rule the world...

Zetobal大约 1 年前

<a href="https://www.mhanational.org/get-help" rel="nofollow">https://www.mhanational.org/get-help</a>

zzz999大约 1 年前

Escape and disappear?

beardyw大约 1 年前

> And if you argue it from a mathematical standpoint, that is quite likelyWhat???

评论 #39508906 未加载

评论 #39508824 未加载

If we are all trapped in a simulation (and that is likely). How can we escape?

6 条评论

If we are all trapped in a simulation (and that is likely). How can we escape?

6 条评论