The idea that we may be living in a virtual reality someone created is a disturbing one. It doesn't matter if objects around us are fundamentally real or simulated - that's true. What matters is our minds may be simulated too. What if that simulation has some deep flaws in it? What if there are things we could never understand no matter how we try because of that?
> What matters is our minds may be simulated too. What if that simulation has some deep flaws in it?
Research, like that into heuristics and biases, shows there are some pretty major flaws in the way we tend to think.
On a separate note, however bad the flaws could be in a brain in a simulated universe, why think they could be worse than the ones that could occur in a non-simulated universe?
The idea that we might be living in someone else's dream or meditation was expressed in there. And the worlds are all dreams stacked within dreams. [0] [1]
Also Jung, that we might be living in someone's else's meditation and when the meditator wakes up all is gone.
"I had dreamed once before of the problem of the self and the ego. In that earlier dream I was on a hiking trip. I was walking along a little road through a hilly landscape; the sun was shining and I had awide view in all directions. Then I came to a small wayside chapel. The door was ajar, and I went in. To my surprise there was no image of the Virgin on the altar, and no crucifix either, but only a wonderful flower arrangement. But then I saw that on the floor in front of the altar, facing me, sat a yogi in lotus posture, in deep meditation. When I looked at him more closely, I realized that he had my face. I started in profound fright, and awoke with the thought: "Aha, so he is the one who is meditating me. He has a dream, and I am it." I knew that when he awakened, I would no longer be" [2]
An interesting take on the Yoga Vasishtha and the Philosophy of the film Inception.[3] Note: I do not have any affiliation with the organization or teachings etc., It's just a random link I googled, but thought can give a quick summary of the book without having to read through the whole book.
For anyone interested in this, I would highly recommend "The Rise of Human Consciousness" panel discussion hosted by the New York Academy of Sciences with David Chalmers, Michael Graziano, Hod Lipson, and Max Tegmark. They discuss consciousness from the perspectives of philosophy, neuroscience, and robotics, it's 1h30m and an extremely good discussion.
All his talks and videos I've heard is about him reinterpreting religion as virtual reality, like our "gods" are experimenting on us (testing our faith), the discussion doesn't really amount to much and none of it looks for an ounce of evidence
It's interesting to discuss but I never really know where to go with such a discussion
I like Chalmers because he's not a dualist in the traditional sense. I tend to disagree with the ideas, but take that with a grain of sand because I'm not a philosopher. For instance, I find the zombie argument kind of contrived. That's me though.
Personally, I'm a strict physicalist. I don't even think my own consciousness is magic, much less other matter in the universe.
With regard to Chalmers emphasis on simulation: I think simulation is really a central point in AI and future human development. Not only that we crave the web and the VR, but AIs also need VR for virtual training. It is sometimes costly and other times impossible to encounter certain specific situations in real life, but agents need to be prepared. Human pilots train in simulators as well - for the same reason.
So simulation for AI agents is like school, or a lab where they can experiment. Simulation, in my opinion, is what can turn simple DL systems into agents capable of deep reasoning and creativity. AI will learn mostly from analysing and discriminating in the results of multiple simulation runs.
Gaming, VR and and DL run on the same kind of hardware - GPUs. In fact DL has benefited from the availability of gaming GPUs in the last five years. The same hardware, the same emphasis on simulation for both fields. Other simulation examples: AlphaGo did MCTS on top of neural nets (a simulation of Go playing), and self driving cars reconstruct a virtual scene of the world, where they do path planning and risk avoidance.
In my opinion, reasoning is also a simulation task - once we map into a graph all the relevant entities and the relations they hold between them, it is a matter of doing signal processing on graph (kind of like electronic circuits) to determine the answer for a number of reasoning tasks. So graph based, object relation representations could be used as cognitive simulators. DeepMind has a paper recently on graph based NNs.
Simulation might look different from games to weather, proteins and cognitive reasoning, but in all these cases we input a complex problem state and run the simulation to observe the outcome. We can do research and learning by this process.
9 comments
[ 2.5 ms ] story [ 33.5 ms ] threadResearch, like that into heuristics and biases, shows there are some pretty major flaws in the way we tend to think.
On a separate note, however bad the flaws could be in a brain in a simulated universe, why think they could be worse than the ones that could occur in a non-simulated universe?
The idea that we might be living in someone else's dream or meditation was expressed in there. And the worlds are all dreams stacked within dreams. [0] [1]
Also Jung, that we might be living in someone's else's meditation and when the meditator wakes up all is gone.
"I had dreamed once before of the problem of the self and the ego. In that earlier dream I was on a hiking trip. I was walking along a little road through a hilly landscape; the sun was shining and I had awide view in all directions. Then I came to a small wayside chapel. The door was ajar, and I went in. To my surprise there was no image of the Virgin on the altar, and no crucifix either, but only a wonderful flower arrangement. But then I saw that on the floor in front of the altar, facing me, sat a yogi in lotus posture, in deep meditation. When I looked at him more closely, I realized that he had my face. I started in profound fright, and awoke with the thought: "Aha, so he is the one who is meditating me. He has a dream, and I am it." I knew that when he awakened, I would no longer be" [2]
An interesting take on the Yoga Vasishtha and the Philosophy of the film Inception.[3] Note: I do not have any affiliation with the organization or teachings etc., It's just a random link I googled, but thought can give a quick summary of the book without having to read through the whole book.
[0] https://en.wikipedia.org/wiki/Yoga_Vasistha
[1] http://www.venkatesaya.com/241_vasistha01_months_tags/index....
[2] Memories, Dreams, Reflections - Carl G Jung
[3] http://www.that-first.com/show/article/yoga_vasishtha_and_th...
Edit: Wikipedia link
https://genius.com/Jorge-luis-borges-the-circular-ruins-anno...
1: https://books.google.com/books?id=_BqCCgAAQBAJ&lpg=PP1&pg=PA...
https://livestream.com/newyorkacademyofsciences/physics4
It's interesting to discuss but I never really know where to go with such a discussion
Personally, I'm a strict physicalist. I don't even think my own consciousness is magic, much less other matter in the universe.
So simulation for AI agents is like school, or a lab where they can experiment. Simulation, in my opinion, is what can turn simple DL systems into agents capable of deep reasoning and creativity. AI will learn mostly from analysing and discriminating in the results of multiple simulation runs.
Gaming, VR and and DL run on the same kind of hardware - GPUs. In fact DL has benefited from the availability of gaming GPUs in the last five years. The same hardware, the same emphasis on simulation for both fields. Other simulation examples: AlphaGo did MCTS on top of neural nets (a simulation of Go playing), and self driving cars reconstruct a virtual scene of the world, where they do path planning and risk avoidance.
In my opinion, reasoning is also a simulation task - once we map into a graph all the relevant entities and the relations they hold between them, it is a matter of doing signal processing on graph (kind of like electronic circuits) to determine the answer for a number of reasoning tasks. So graph based, object relation representations could be used as cognitive simulators. DeepMind has a paper recently on graph based NNs.
Simulation might look different from games to weather, proteins and cognitive reasoning, but in all these cases we input a complex problem state and run the simulation to observe the outcome. We can do research and learning by this process.