This lecture covers DNA and the origin of life on Earth, the Fermi Paradox (is there alien life?), AI and its implications for the Simulation Question: Could our universe be a simulation? Are we machines, but don't know it?
Steve talks with Skype founder and global tech investor Jaan Tallinn. Will the coronavirus pandemic lead to better planning for future global risks? Jaan gives his list of top existential risks and describes his efforts to call attention to AI risk. They discuss AGI, the Simulation Question, the Fermi Paradox and how these are all connected. Do we live in a simulation of a quantum multiverse?
In mathematics, a manifold is a topological space that locally resembles Euclidean space near each point.
Steve Hsu and Corey Washington have been friends for almost 30 years, and between them hold PhDs in Neuroscience, Philosophy, and Theoretical Physics. Join them for wide ranging and unfiltered conversations with leading writers, scientists, technologists, academics, entrepreneurs, investors, and more.
Steve Hsu is VP for Research and Professor of Theoretical Physics at Michigan State University. He is also a researcher in computational genomics and founder of several Silicon Valley startups, ranging from information security to biotech. Educated at Caltech and Berkeley, he was a Harvard Junior Fellow and held faculty positions at Yale and the University of Oregon before joining MSU.
Corey Washington is Director of Analytics in the Office of Research and Innovation at Michigan State University. He was educated at Amherst College and MIT before receiving a PhD in Philosophy from Stanford and a PhD in a Neuroscience from Columbia. He held faculty positions at the University Washington and the University of Maryland. Prior to MSU, Corey worked as a biotech consultant and is founder of a medical diagnostics startup.
The Wuhan coronavirus outbreak (see WHO resource) caused me to look back at a paper I wrote in 2003 with A. Zee. We were motivated at the time by the recent SARS outbreak. Some results in the paper may be relevant today.
For Wuhan coronovirus the important parameters such as R0 (average number of secondary cases caused by a single infected individual) and lethality are still to be determined.
We develop simple models for the global spread of infectious diseases, emphasizing human mobility via air travel and the variation of public health infrastructure from region to region. We derive formulas relating the total and peak number of infections in two countries to the rate of travel between them and their respective epidemiological parameters.
From the conclusions:
One interesting conclusion from our models is that typical international mobility – the probability per unit time of international travel for a given infected individual, estimated at mi→j ∼ 10−5 per week – is still sufficiently small that a country with well-developed public health infrastructure (effectively, a negative eigenvalue λ) can resist an epidemic even when other more populous countries experience complete saturation. In the quasi-realistic simulation 1 (figures (1),(2)), of order 10^5 infections occur in country 2, even though the disease has swept completely through country 1. In reaching this conclusion, we kept the mobility parameter fixed during the outbreak, and did not assume any draconian quarantine on international travelers arriving in country 2. Such measures would reduce the number of infections in country 2 considerably. Of course, this conclusion assumes that the public health infrastructure in country 2 remains robust during the outbreak. In the nonlinear simulation 3 (figures (6), (7)), we see that a breakdown in the medical system can lead to grave consequences.
Great talk on the use of game engines (virtual worlds) for AI/ML agent training. Even if you are already knowledgeable about this topic, the examples he shows will be useful to guide your intuition as to what is possible, what is easy/hard with current technology and methods. Don't miss the puppies :-)
Conceptually, I would say there is not much new since the early successes with simpler (e.g., Atari) games. See papers/talks by Schmidhuber in this 2014 post. IIRC, the concept of curiosity: seeking "surprise" = large chunks of information = large model updates, was formulated already some time ago.
One thing that is new is the use of physics engines in the virtual worlds - i.e., the AI has to deal with dynamics as in the real world. It seems to me that routine task automation, such as in manufacturing, is not that much harder than what is being done here in game worlds with good physics engines. (Note I'm not referring to the mechanical engineering or physical robotics challenges, which could be significant, just the ML part.) Replacement of humans in many routine tasks seems now a matter of economics tradeoffs and application of known technologies rather than big breakthroughs.
I've always thought we'd get to AGI after consuming a lot of FLOPS training agents in increasingly realistic virtual worlds. Of course, this makes one wonder whether we ourselves exist in a simulation ;-)
The Simulation Hypothesis is the idea that our universe might be part of a simulation: we are not living in base reality. (See, e.g., earlier discussion here.)
There are many versions of the argument supporting this hypothesis, which has become more plausible (or at least more popular) over time as computational power, and our familiarity with computers and virtual worlds within them, has increased.
Modern cosmology suggests that our universe, our galaxy, and our solar system, have billions of years ahead of them, during which our civilization (currently only ~10ky old!), and others, will continue to evolve. It seems reasonable that technology and science will continue to advance, delivering ever more advanced computational platforms. Within these platforms it is likely that quasi-realistic simulations, of our world, or of imagined worlds (e.g., games), will be created, many populated by AI agents or avatars. The number of simulated beings could eventually be much larger than the number of biologically evolved sentient beings. Under these assumptions, it is not implausible that we ourselves are actually simulated beings, and that our world is not base reality.
One could object to using knowledge about our (hypothetically) simulated world to reason about base reality. However, the one universe that we have direct observational contact with seems to permit the construction of virtual worlds with large populations of sentient beings. While our simulation may not be entirely representative of base reality, it nevertheless may offer some clues as to what is going on "outside"!
The simulation idea is very old. It is almost as old as computers themselves. However, general awareness of the argument has increased significantly, particularly in the last decade. It has entered the popular consciousness, transcending its origins in the esoteric musings of a few scientists and science fiction authors.
The concept of a quantum computer is relatively recent -- one can trace the idea back to Richard Feynman's early-1980s Caltech course: Physical Limits to Computation. Although quantum computing has become a buzzy part of the current hype cycle, very few people have any deep understanding of what a quantum computer actually is, and why it is different from a classical computer. A prerequisite for this understanding is a grasp of both the physical and mathematical aspects of quantum mechanics, which very few possess. Individuals who really understand quantum computing tend to have backgrounds in theoretical physics, physics, or perhaps computer science or mathematics.
The possibility of quantum computers requires that we reformulate the Simulation Hypothesis in an important way. If one is willing to posit future computers of gigantic power and complexity, why not quantum computers of arbitrary power? And why not simulations which run on these quantum computers, making use of quantum algorithms? After all, it was Feynman's pioneering observation that certain aspects of the quantum world (our world!) are more efficiently simulated using a quantum computer than a classical (e.g., Turing) machine. (See quantum extension of the Church-Turing thesis.) Hence the original Simulation Hypothesis should be modified to the Quantum Simulation Hypothesis: Do we live in a quantum simulation?
There is an important consequence for those living in a quantum simulation: they exist in a quantum multiverse. That is, in the (simulated) universe, the Many Worlds description of quantum mechanics is realized. (It may also be realized in base reality, but that is another issue...) Within the simulation, macroscopic, semiclassical brains perceive only one branch of the almost infinite number of decoherent branches of the multiverse. But all branches are realized in the execution of the unitary algorithm running on qubits. The power of quantum computing, and the difficulty of its realization, both derive from the requirement that entanglement and superposition be maintained in execution.
Given sufficiently powerful tools, the beings in the simulation could test whether quantum evolution of qubits under their control is unitary, thereby verifying the absence of non-unitary wavefunction collapse, and the existence of other branches (see, e.g., Deutsch 1986).
We can give an anthropic version of the argument as follows.
1. The physical laws and cosmological conditions of our universe seem to permit the construction of large numbers of virtual worlds containing sentient beings.
2. These simulations could run on quantum computers, and in fact if the universe being simulated obeys the laws of quantum physics, the hardware of choice is a quantum computer. (Perhaps the simulation must be run on a quantum computer!)
If one accepts points 1 and 2 as plausible, then: Conditional on the existence of sentient beings who have discovered quantum physics (i.e., us), the world around them is likely to be a simulation running on a quantum computer. Furthermore, these beings exist on a branch of the quantum multiverse realized in the quantum computer, obeying the rules of Many Worlds quantum mechanics. The other branches must be there, realized in the unitary algorithm running on (e.g., base reality) qubits.
Seems like Elon might have been high before the interview even started 8-) Early discussion focused on AI, Neuralink, Singularity risk, etc. Simulation @43min.
Let R = the ratio of number of artificially intelligent virtual beings to the number of "biological" beings (humans). The virtual beings are likely to occupy the increasingly complex virtual worlds created in computer games, like Grand Theft Auto or World of Warcraft (WOW will earn revenues of a billion dollars this year and has millions of players). In the figure below I have plotted the likely behavior of R with time. Currently R is zero, but it seems plausible that it will eventually soar to infinity. (See previous posts on the Singularity.)
If R goes to infinity, we are overwhelmingly likely to be living in a simulation...
... Think of the ratio of orcs, goblins, pimps, superheroes and other intelligent game characters to actual player characters in any MMORPG. In an advanced version, the game characters would themselves be sentient, for that extra dose of realism! Are you a game character, or a player character? :-)
I'm holding off on this in favor of a big binge watch.
Certain AI-related themes have been treated again and again in movies ranging from Blade Runner to the recent Ex Machina (see also this episode of Black Mirror, with Jon Hamm). These artistic explorations help ordinary people think through questions like:
What rights should be accorded to all sentient beings?
Can you trust your memories?
Are you an artificial being created by someone else? (What does "artificial" mean here?)
I suspect Castronova would agree with me that if AI is eventually successful, then many, perhaps almost all, sentient beings in our future lightcone will be game characters who are themselves unaware that they live in a simulation.
