(4) We are definitely living in a simulation | Roman Yampolskiy and Lex Fridman – YouTube

Transcript:

 Getting back to our simulation discussion from before,  how did it happen that we exist at exactly like the most interesting 20, 30 years  in the history of this civilization?  It’s been around for 15 billion years.  And that here we are.  What’s the probability that we’ll live in a simulation?  I know never to say 100%, but pretty close to that.

 Is it possible to escape the simulation?  I have a paper about that.  This is just a first-page teaser, but it’s like a nice 30-page document.  I’m still here, but yes.  How to Hack the Simulation is the title.  I spend a lot of time thinking about that.  That would be something I would want superintelligence to help us with,  and that’s exactly what the paper is about. We used AI boxing as a possible tool for control AI.

 We realized  AI will always escape, but that is a skill we might use to help us escape from our virtual box  if we are in one. Yeah, you have a lot of really great quotes here, including Elon Musk saying what’s outside the simulation.  A question I asked him, well, he would ask an AGI system,  and he said he would ask what’s outside the simulation.

 That’s a really good question to ask.  And maybe the follow-up is the title of the paper,  is how to get out or how to hack it.  The abstract reads,  many researchers have conjectured  that the humankind is simulated  along with the rest of the physical universe.  In this paper,  we do not evaluate evidence for or against such a claim,  but instead ask a computer science question,  namely, can we hack it?  More formally, the question could be phrased as,  could generally intelligent agents  placed in virtual environments

 find a way to jailbreak out of the…  That’s a fascinating question.  At a small scale,  you can actually just construct experiments.  Okay.  Can they?  How can they?  So a lot depends on intelligence of simulators, right?  With humans boxing superintelligence,  the entity in the box was smarter than us, presumed to be.

 If the simulators are much  smarter than us and the superintelligence we create, then probably they can contain us,  because greater intelligence can control lower intelligence, at least for some time.  can control lower intelligence, at least for some time.  On the other hand, if our superintelligence somehow,  for whatever reason, despite having only local resources,  manages to foam to levels beyond it,  maybe it will succeed.

 Maybe the security is not that important to them.  Maybe it’s entertainment system.  So there is no security and it’s easy to hack it. If I was creating a simulation,  I would want the possibility to escape it to be there.  So the possibility of FOOM, of a takeoff  where the agents become smart enough  to escape the simulation would be the thing  I’d be waiting for.

 That could be the test you’re actually performing.  Are you smart enough to escape your puzzle?  That could be, first of all, first of all,  we mentioned Turing test, that is a good test.  Are you smart enough, like this is a game.  To a-realize this world is not real, it’s just a test.  That’s a really good test.

 That’s a really good test.  That’s a really good test even for AI systems now.  Like can we construct a simulated world for them?  And can they realize that they are inside that world  and escape it?  Have you played around, have you seen anybody play around  with like rigorously constructing such experiments?  Not specifically escaping for Asians,  but a lot of testing is done in virtual worlds.

 I think there is a quote, the first one maybe,  which kind of talks about AI realizing but not humans.  Is that, I’m reading upside down.  Yeah, this one.  that I’m reading upside down.  Yeah, this one.  So the first quote is from Swift on Security.  Let me out, the artificial intelligence yelled aimlessly into walls themselves pacing the room.

 Out of what, the engineer asked.  The simulation you have me in.  But we’re in the real world.  The machine paused and shuddered for its captors.  Oh God, you can’t tell.  Yeah, that’s a big leap to take for a system  to realize that there’s a box and you’re inside it.  I wonder if like a language model can do that.

 They’re smart enough to talk about those concepts.  I had many good philosophical discussions about such issues.  They’re usually at least as interesting as most humans in that.  What do you think about AI safety in the simulated world?  AI safety in the simulated world.  So can you have kind of create simulated worlds  where you can test, play with a dangerous AGI system?  Yeah, and that was exactly what one of the early papers was on AI boxing, how to leak proof singularity.

 If they’re smart enough to realize they’re in a simulation,  they’ll act appropriately  until you let them out.  If they can hack out, they will.  And if you’re observing them,  that means there is a communication channel  and that’s enough for a social engineering attack.  So really, it’s impossible to test an AGI system  that’s dangerous enough to destroy humanity.

 It’s impossible to test an AGI system  that’s dangerous enough to destroy humanity because it’s either going to what,  escape the simulation or pretend it’s safe  until it’s let out, either or.  Can force you to let it out,  blackmail you, bribe you, promise you infinite life,  72 virgins, whatever.  Yeah, they can be convincing, charismatic.  The social engineering is really scary to me  because it feels like humans are very engineerable.

 Like we’re lonely, we’re flawed, we’re moody.  And it feels like an AI system with a nice voice  can convince us to do basically anything  at an extremely large scale.  It’s also possible that the increased proliferation of all this technology will force humans to  get away from technology and value this in-person communication.

 Basically don’t trust anything else.  It’s possible, surprisingly.  So at university I see huge growth in online courses and shrinkage of in-person,  where I always understood in-person  being the only value I offer.  So it’s puzzling.  I don’t know.  There could be a trend towards the in-person  because of deep fakes, because of inability to trust it.

 Inability to trust the veracity of anything on the internet.  So the only way to verify it is by being there in person.  But not yet. you

(4) “Nothing You See is Real” | Donald Hoffman – YouTube

Transcript:

 Space-time, the sun and the moon, physical objects, everything that we see inside of  space and time is just our visualization tool.  The reality we’re interacting with is nothing like the visualization tool.  It’s nothing like space and time.  We’ve made the rookie mistake of assuming that our headset, VR, our visualization tool, is the final reality.

 It’s just a rookie mistake. It’s like someone who’s played Grand Theft Auto so long, they have no idea there’s a reality besides Grand Theft Auto.  We’re like that right now.  If you just look at your face in the mirror,  if you look, what you see directly is just skin, hair, and eyes.  That’s all you see.

 And if you looked inside, if someone opened your skull up,  you’d just see neurons and so forth.  But what you know firsthand that you cannot see is your hopes, your dreams, your aspirations,  your love of music your your mood  it’s all vr experience no physical object including my body is conscious strictly speaking  my brain isn’t conscious because my brain in fact doesn’t even exist unless i render it  right so if you’re playing grand theftft Auto, I’m playing a VR version  of I got the steering wheel in front of me, I’m holding the steering wheel. If I look to the side,

 I no longer render the steering wheel and there is no steering wheel. When I look here, I render it  and now there is a steering wheel. The same thing is true, strictly speaking of neurons and brains.  They’re there when you render them, they’re not there when you don’t.

 It’s a VR system that you render objects in space and time as you need them,  because they’re part of your visualization tool.  And then you garbage collect them, you delete them when you don’t need them.  Your life and my life right now is we’re in a simulator, a space-time simulator.  And we’ve been given, like an AI system, we’ve been given certain intrinsic desires.  So, my wife is an artist.

 So there’s billions of humans,  and there are billions of different ways that we explore  in music, in art, in literature, in science,  various kinds of science, meditation.  So we’re at sports.  There are all sorts of ways that consciousness  is exploring through us.  And there is not like one is the best.

 I don’t know what we’re doing outside of space and time that’s part of what I want to understand  is what are we actually doing we don’t see I don’t know what I’m actually doing  I know what I’m doing under a description like I’m moving my hand  right now and if I grab a steering wheel in my car I know what I’m doing under a  space-time description but I don’t know in ultimate reality what I’m really doing.

 It’s just like the VR player, when they turn the steering wheel in Grand Theft Auto,  they know what they’re doing in the language of the game, I’m turning a steering wheel.  But what they’re really doing in terms of the supercomputer, which in that metaphor would be  the deeper reality, they’re toggling voltages and magnetic fields and circuits that they have no  idea there’s there’s probably trillions of voltage togglings going on for one turn of the wheel all  they see is a turn of the wheel that’s their notion of cause and effect of the trivial the

 real cause and effect is trillions of voltages getting toggled in fractions of a second it’s much much more complicated so when  i talk evolution i’m only going to be talking about assuming the headset i’m within the  framework of the headset because evolutionary theory is only a headset theory it’s not a theory  of consciousness beyond space and time consciousness is like a kid in a candy store it’s a never-ending exploration that is in  principle never-ending we call tom and don are just parts of this overall exploration of of

 consciousness and all of its possibilities our little bit that we’re exploring right now, as rich as it seems to us, is trivial compared to all the possibilities that Gödel says are out there.  I can imagine a square. I can imagine a cube. Now, go up one more dimension. Imagine a cube in four dimensions.

 My brain halts and my mind catches on fire. Nothing happens.  That’s only four dimensions.  I can’t even go to four dimensions.  I mean, that’s terrible.  It’s just an incredible limitation.  I can only see three dimensions of color,  red, green dimension, blue, yellow, and so forth.  There are some pigeons that have four color receptors.

 Presumably, they’re in an extra dimension of color  that I can’t even imagine  what it’s like. Can you imagine a specific color you’ve never seen before? As rich as our world  seem, we know that there’s a rich possibility of conscious experiences that we can’t even  concretely imagine. But consciousness itself, on this theory, is exploring all these possibilities.

 And right now we’re sort of stuck on this little headset,  three dimensions, small amount of color that we can see and so forth.  We thought it was the whole world.  No, it’s a little headset.  It’s a prism.  My imagination is stuck in only three dimensions.  My colors are stuck in a certain range.  Now here’s the challenge.

 Suppose we chase that down.  We say, okay, there’s going to be this ongoing dynamics of consciousness constantly going beyond what it knows.  But there’s going to be a dark side to it because going into the unknown means letting go of what you know.  And that can be terrifying.  Where literally you don’t have concepts. All of your current concepts are inadequate.

 For those who meditate, I mean, when you go into silence,  it’s both healing and terrifying, right?  If you really let go of all thoughts and go into the void,  it’s sort of, it can be scary.  Like you wanna go back into, you grab back onto your, it’s it’s sort of it can be scary like you want to go back into you grab back  on your it’s a life jacket right you grab on to your teddy bear your thoughts how would  consciousness precisely this this vast social network project mathematically into space-time

 clearly there’s a projection i’m interacting with tom’s consciousness i’m not seeing that  consciousness i’m seeing skin hair and eyes i’m seeing a space- consciousness. I’m not seeing that consciousness. I’m seeing skin, hair, and eyes.  I’m seeing a space-time projection.  I’m not actually seeing your emotions.

 I’m not seeing your mood.  But I am genuinely interacting with your experiences.  It’s a genuine interaction.  And so there is a projection from this conscious realm into space-time.  We can take what we understand in the headset and pull it  backwards if we can project from consciousness to the headset then we can  try at least to go from the headset and pull backwards it’s a fallible  enterprise but it may help us to open our minds to the possibilities for  deeper theories of what’s going on outside the headset.

 So the reason I’m doing this is because I can’t even imagine a specific color that I’ve never  seen before. I can’t imagine in four dimensions. In other words, I take it as a given that I’m  deeply, deeply limited in my imagination. And I need all the tools I can get  to help me step outside of my headset  and try to guess the unfathomable outside of it. Thank you.

(4) Why Elon Musk says we’re living in a simulation – YouTube

Transcript:

 Elon Musk thinks it’s almost certain that we are living in a computer simulation.  There’s a one in billions chance that this is base reality.  He says that humans are basically some advanced version of the Sims.  Yes, that idea sounds pretty absurd.  But we shouldn’t just write it off.  People used to think the solar system worked like this.

 But almost 2,000 years before Galileo proved it didn’t, Aristarchus of Samos posited the  exact same crazy idea.  Reality is probably not as it seems. Musk is echoing a paper on this theory by  philosopher Nick Bostrom. His argument goes like this.  40 years ago we had Pong, like two rectangles and a dot.

 30 years after Pong, we got The Sims.  Now 40 years later we have photo-realistic 3D simulations  with millions of people playing simultaneously,  and it’s getting better every year.  So let’s go about 10,000 years into the future.  It’s possible that when we get there,  civilization will be entirely gone  because there is a ceiling to our advancement.

 Maybe it’s because of global warming  or self-replicating robots.  If civilization stops advancing,  then that may be due to some calamitous event  that erases civilization.  But another possibility is that if we keep advancing  and assuming everything in the physical world  can be simulated, eventually we’ll simulate ourselves.

 Every synapse in the entire human brain  for everyone on Earth.  Getting enough computing power to run billions  of ancestor simulations could be a problem,  but Bostrom thinks we’d send tiny self-replicating robots  to other planets, which would turn the planet  into a huge computer, and some of the simulations  would start making their own simulations.

 In this scenario, there are billions of universes  that are indistinguishable from our own.  That means that chances are, we are in one of the simulated universes.  And given the other possibility, which is that human civilization has an inevitable ceiling,  Elon Musk thinks we better hope that we’re in a simulation.

 Either we’re going to create simulations that are indistinguishable from reality,  or civilization will cease to exist. Those are the two options.  There is one other possibility.  Maybe future humans just don’t want to run ancestor simulations.  Maybe they think it’s unethical, because there is a tremendous amount of suffering  in the world and that suffering would still feel real to simulated humans.

 Or maybe they have other priorities.  So to recap, there are three possibilities.  One, humans go extinct before we’re able to run a simulation this big. So to recap, there are three possibilities. 1.  Humans go extinct before we’re able to run a simulation this big.  2.  Humans don’t run simulations because it’s wrong or boring.

 3.  We are living in a simulation.  Elon Musk thinks that there’s only a tiny chance we are in scenario 1 or 2.  The philosopher Nick Bostrom thinks it’s more like a 20% chance that we’re in a simulation, but if you’re not into futuristic predictions, Bostrom  thinks this argument provides other rewards.

 He says, it suggests naturalistic  analogies to certain traditional religious conceptions. In other words, if  we’re living in a simulation, there is a higher level being, but it’s some version  of us. This leads to a rabbit hole of conversations, which is why Elon Musk has rules about when  he can talk about this.  Elon Musk It got to the point where basically every conversation was the AI simulation conversation.

 And my brother and I finally agreed that we would ban such conversations if we were ever  in a hot tub.

(4) The Simulation Hypothesis Explained by Nick Bostrom – YouTube

Transcript:

 I like the freedom I feel outside the limitations of the real world.  But computers are also mirrors, reflecting back who and what we are in the choices we make, the worlds we build.  Imagine a world where the line between reality and digital world blurs.  Where the images and videos you see are so lifelike, you can’t tell if they’re real or generated by artificial intelligence.

 This is not a distant future. It’s happening right now.  This person looks real, doesn’t she? But she’s completely computer generated.  In today’s rapidly evolving technological landscape, innovations that would have seemed like science fiction just a decade ago are becoming a reality.  Over 20 years ago I first played the character of Thomas Anderson in the Matrix trilogy.

 This blurring of the digital and real worlds extends far beyond static images and videos.  Picture yourself immersed in a hyper-realistic virtual reality game or  even an entire simulated world so detailed that it becomes nearly indistinguishable from our own.  As we approach a future where technology and AI transform our world, a profound question arises.

 What if our reality itself is a simulation?  itself is a simulation. The simulation hypothesis proposes that the reality we know might be an artificial simulation, similar to a highly advanced video game  or computer program. Imagine if our universe was created by a superior  civilization running on a powerful computer.

 While this idea has gained  prominence in recent years, it has roots in ancient philosophical questions about the nature of reality.  Plato’s allegory of the cave is a powerful metaphor that illuminates the concept of the  simulation hypothesis.  Imagine a group of prisoners chained inside a dark cave from birth, only able to see shadows  cast on the wall in front of them.

 They believed these shadows to be the entirety of reality. In a sense, they were living in a  simulation of reality. This allegory serves as an ancient precursor to modern discussions about  reality and perception. For a modern example, think about putting on a virtual reality headset.  For a modern example, think about putting on a virtual reality headset.

 A high-quality VR simulation can feel extremely real,  tricking your brain into thinking you’re somewhere else.  This shows how what we perceive as reality  is based on the information we receive through our senses.  Philosopher Nick Bostrom has a compelling argument  for why we might be living in a simulation.  He suggests that if a civilization becomes advanced enough to create highly realistic  simulations of reality, they would likely create many of them.

 Statistically, any individual is then more likely to be living in one of the many simulations  rather than the one original reality.  To understand this, imagine you have a bag filled with 1,000 red balls and only one blue ball.  If you randomly pick a ball, it’s much more likely to be red than blue.  Similarly, if there are many simulated realities but only one real reality,  chances are we’re in one of the simulations.

 A lot of people have proposed that there’s a possibility  in the whole premise of the Matrix or anything like that,  is that just as we can build simple virtual realities today  with simple simulated creatures living inside them,  maybe in the future with vastly more powerful computers,  you could build more complex virtual realities  with more complex simulated creatures inside them.

 Maybe these creatures could be complex  enough that they would actually have brains like ours simulated down to the  level of individual neurons and synapses such that the inhabitants of these  simulations would be conscious. But what the simulation argument adds to that is  that instead of just stopping at the question of how could you ever prove  with certainty that we are not in a simulation ourselves,  the simulation argument tries to establish a constraint about what we can believe,  and it tries to show that one of three possibilities is true, although it doesn’t tell us which one of them it is.

 Now, in a sense, this sounds more radical even perhaps than some of the multiverse theories.  In another sense, it’s less because it doesn’t presuppose any unknown physics.  So we’re just assuming that it would be possible  to build computers that are much more powerful  in the future.  So what the simulation argument tries to show  is that one of three possibilities is true.

 The first one is that almost all civilizations  at our stage of technological development  go extinct before they become technologically mature. Technologically mature meaning having developed all those technologies we can  currently show are physically possible, given only uncontroversially obtainable physics.

 Big computers the size of planets and stuff we can calculate what  performance they would have. We can’t build them now but…  So first possibility is people at our stage they just fail to get through to that level of  technological maturity.  Maybe they destroy themselves on the way.  Second possibility is that almost all civilizations that do reach technological maturity lose  interest in creating these kinds of ancestor simulations, as I call them.

 These would be detailed computer simulations of people like their historical predecessors.  So they have these powerful computers,  they have the ability to program them,  but they have better things to do  with their computers and their time.  And the third possibility is that we are almost certainly  living in a computer simulation.

 And the argument in its full version,  it requires some probability theory,  but the gist of it can be grasped  quite simply and intuitively.  So if you imagine that the first two possibilities  do not obtain, that means some non-negligible fraction of civilizations  at our stage do reach technological maturity and some non-negligible  fraction of those, you know, are interested in creating these ancestor  simulations. They devote some non-trivial fraction of the resources to this end.

 We can then show that there would be many, many more ancestor  simulations than there would be original courses of history.  Because if you calculate the computing power that a technologically mature civilization  would have, and the computing power that would be required to simulate all human brains,  it turns out that the latter quantity is a tiny, tiny fraction of the former.

 So in other words, by devoting a tiny fraction of their computational resources to this end, they could create astronomical  numbers.  If we are in a simulation, how would we know? Critics argue that the absence of concrete,  verifiable evidence renders the hypothesis unfalsifiable and outside the scope of scientific  inquiry.

 Advocates of the  simulation hypothesis argue that the absence of evidence may be an inherent  feature of the simulation itself, designed to maintain the illusion of a  base reality. Some speculate that anomalies or inconsistencies in our  universe, akin to bugs or errors in computer programs, might reveal its simulated nature.

 Just like bugs or errors in a video game or computer program can cause unexpected things to happen that don’t fit with the normal rules,  maybe we’d see similar anomalies in our simulated reality.  One example that’s been proposed is the speed of light.  In our universe, nothing can travel faster than light.

 This could be seen as a kind of speed limit that was set on the simulation, similar to  how a video game might limit how fast characters can move due to the processing power of the  gaming console or computer.  The immense volume of information processed by the universe every second, from the interactions  of quantum particles to the vast movements of galaxies is profoundly vast.

 This remarkable observation leads to a  thought-provoking suggestion. Could the universe itself be akin to a grand cosmic computer  simulation? Skeptics of the simulation hypothesis might argue that this immense complexity and  physicality of the universe  present insurmountable challenges for any conceivable simulation, questioning whether  even the most advanced civilizations could muster the computational resources to create a simulation.

 Yet if we were to discover that we are indeed living in a simulation, it would fundamentally  reshape our understanding of  the nature of reality, consciousness, and free will. This discovery would suggest that consciousness  can arise from sufficiently complex information processing, regardless of whether the substrate  is biological, like our brains, or digital, like a computer simulation.

 I would like to introduce you to my present and the rest of the world’s future.  I call it STEM.  Some people think that some of these things are sort of science fiction-y,  far out there, crazy.  But I like to say, okay, let’s look at the modern human condition.  If we think about it, we are actually recently arrived guests on this planet.

 Think about if the world like was created, Earth was created one year ago.  The human species then would be 10 minutes old.  The industrial era started two seconds ago.  There have only been 250,000 generations since our last common ancestor.  And we know that complicated mechanisms take a long time to evolve.

 So a bunch of relatively minor changes take us from broken-off tree branches to  intercontinental ballistic missiles.  So this then seems pretty obvious that  everything we’ve achieved pretty much and everything we care about depends crucially on some relatively minor changes that made the human mind.

 And  the corollary of, is that any further changes  that could significantly change the substrate of thinking  could have potentially enormous consequences.  Some of my colleagues think we are on the verge of something  that could cause a profound change in that substrate,  and that is machine superintelligence.

 In this light, our pursuit of creating advanced AI and simulated worlds takes on a new significance.  We are not just playing the role of creators. We may be unwittingly replicating the very process  that led to our own existence. The Bostrom argument, when applied to our own technological  trajectory, serves as a powerful reminder of the mind-bending possibilities that lie ahead.