Looking at that map of representations of the DNQ agent playing Space Invaders I can't help thinking if it really has learned any concepts that are similar to what a human would learn. Does the DNQ agent have the concepts of ship, shield, bullet or space invader? Does it have anything that corresponds to the concept of a group of space invaders? Can it generalize? I'm sure human players could quickly adapt if we changed the game so that the ship would shoot from the top to the bottom instead. Does the DNQ agent have anything analogous to an inner simulator? If we showed it a movie where the ship would fly up to the invaders and collide what would it predict happens next?
My gut feeling is that artificial agents are still far away from having reusable and generalizable concepts. It's one thing, although an impressive one, to use the same framework with identical parameters for different DNQ agents learning different games than it is to use one framework for one agent that learns to play all the games and abstract concepts across them.
I think we have to be really careful when bringing AIXI/Kolmogorov complexity to the discussion. My strong intuition is that AIXI is specifically the type of alien intelligence we want to avoid here. Most programming languages have very short programs that behave in wild and complex ways when analyzed from a human perspective. To an AIXI mind based on those programming languages this kind of behavior is by definition simple and preferred by Occam's Razor. The generalizations to n+1 dimensions that this kind of an intelligence makes are almost guaranteed to be incomprehensible and totally alien to a human mind.
I'm not saying that information theory is a bad approach in itself. I'm just saying that we have to be really careful what kind of primitives we want to give our AI to reason with. It's important to remember that even though human concepts are only a constant factor away from concepts based on Turing machines, those constant factors are huge and the behavior and reasoning of an information theoretic mind are always exponentially dominated by what it finds simple in it's own terms.
The venue has changed! We are now meeting up at Cafe Vanha at Mannerheimintie 3, Helsinki.
It took me couple of years to get to the point where depression started to dominate so I guess a lot of the groundwork was already done. It did really only take me only a couple of months of regular practice to get the fruits of DavidM's method. The heaviest bouts of depression, boredom and insanity occur during meditation but some of that "leaks" to the daily life as well. The really weird stuff I had like a visualisation of a creature that chops of your body parts and dances around with them laughing only occur in meditation. There are no hallucinations outside of meditation and even in meditation you mostly have control of the visualisations and know that they are not real or important. It's more about letting go and the mind sometimes does crazy stuff before settling down. I would be surprised if meditators in general didn't have crazy episodes from time to time. Talking about them is taboo in most traditions. It does make sense because the mind wanders around and picks up stuff associated with meditation. If you hear or read a lot about strange stuff that can appear in meditation your mind will start playing around with the ideas and then you have sort of imagined visualisations that are doubly irrelevant. My issue here is that when people start to practice meditation alone all of the insane stuff can be quite frightening if you don't have a teacher or peers who can guide one through them. So for the lone meditators it's good to know a little about the weird stuff so they don't give them too much attention or freak out.
I'm glad you liked the key-frame metaphor. I'm planning on writing a proper article about these experiences and developing a coherent language that makes sense to other physical algorithms trying to experience themselves as such. However I feel that I need to meditate a little longer to make sure I'm not making up too much nonsense.
The main difference between knowing that the self is an illusion and feeling it as such is control of the self-alief. Even without getting into a meditative state I can steer my experience from the habitual one that I use to socialize and become a flow of information between different conscious and semi-conscious modules. I cannot fully eliminate the experience of my habitual identity but I can see it jump between different modalities while still claiming to be the same agent that it was before jumping. It's hard to believe in an unified self when you can see that the self-experience is just a way for the mind to control itself.
The disbelief becomes even stronger in meditation when the identity is simply not there even though the mind is clearly experiencing something and recording memories.
Losing your identity is not so bad. You can still keep it and be your habitual self whenever the wish to do so arises. This whole thing is pretty great but I still advice making sure that you really want to do this. There is hardly any way of going back.
Let me address some fears associated with losing oneself:
There is great joy in becoming the experience of happiness itself. Bliss needs no target to feel good.
Just as there is joy there is pain. It simply has less chance of sticking anywhere because the sufferer can be toned down at will. To find meaning I recommend focusing on the pain of others and having compassion. They are made of the same bits as you are and there is no fundamental separation. (Compassion is not strictly required if it feels too touchy-feely but where's the harm if done in moderation.)
The autobiographical self won't disappear but the memories won't seem any different from any other autobiographical story. You won't be special for yourself but you will be and feel special in the sense that you'll be writing new chapters in the story. The achievements in your autobiography will be seen as events caused by inner and outer forces. Acknowledging the merit of others in your finest moments and feeling grateful is a wholesome experience. In my opinion feeling inclusive pride (aren't we all great) is superior to feeling exclusive pride (I'm great, aren't I) but do consider that feeling exclusive pride will feel like a bad joke. Luckily there just won't be anyone left to get hurt by that joke. If you are motivated by exclusive pride you can take comfort in the fact that other people will think you're great even after enlightenment. Freeing up the potential of the information processor under your skin will look like "you" doing all kinds of great stuff.
Different things usually do have different frequencies but that depends on the meditative stage I'm in. Stage 2 usually involves different sensations bursting to the foreground of attention vibrating at their own pace and maybe even accelerating in frequency as they appear and decelerating when fading out. Stage 3 is a cacophony of subtle sensations all vibrating at different frequencies and slowly shifting about. Because everything is constantly fading out and stays in the background it's hard to pick up any individual vibrations at all. Trying to "hear" it all at once takes effort but is doable. Now that I no longer experience mind numbing agony in stage 3 it's actually quite an impressive ocean of dark tones. In stage 4 however all vibrations line up and everything that arises vibrates at the same stable frequency as everything else. Even the experience of space beats away at this global rhythm. All of the previously described vibrations are in the 2 to 6 Hz range but there are higher vibrations associated with more precise sensing like feeling the individual hairs on my eyebrow. The experience of the lower frequency vibrations vibrates too but at a higher pace. Another high frequency object is the emptiness between individual moments of conventional space at stage 4. Don't ask me where this space beyond space vibrates into. I have no idea and luckily subjective experiences are not required to add up to a coherent system.
For type 2, this is how things are all the time. The mind is good at hiding the the details and discontinuities in its function and it takes a good look to see how things don't always line up. While the actual discontinuities are hard to notice the periodic jostling around of attention might be easier to pick up. Attention is constantly pushing away and trying to find novel things. I can sort of feel every push when fixing my focus on a particular object. The frequency changes when I engage in an activity that requires more attention but it's hard to pay attention to attention when I'm supposed to be paying attention to the activity. This difficulty of paying attention to attention may even mean that the discontinuity of experience and the subtle pushes of attention are separate phenomena. I just haven't been able to experience them individually.
Stage 3 was awful for me. Mainly because I gave too much weight to the unpleasant experiences. When you're blaming yourself for all the pain pouring out of your subconsciousness you're just repeating the cycle and training the mind to feel bad for feeling bad. As I also stopped meditating this cycle of self-blame solidified into an automatic habit. Neither did it not help that it feels good to wallow in self-pity when you think you deserve it. Needless to say I was depressed, short on energy and life felt meaningless and void of joy. It was hard facing this undercurrent of agony face-on and painful to let go of the solid tracks of negative feedback. It's a challenge not trying to blame yourself for failing not blaming yourself...
But now the Dark Night has given way to a dawn of serene contentment. Even when not meditating I feel this undercurrent of peace and when I'm relaxed I simply feel myself fading in and out of space to a cosmic rhythm. (Not to imply that such a cosmic rhythm actually exists out there or anything like that but subjectively I feel like I'm being caressed by the Universe with every breath I take.)
I perceive two kinds of vibrations. The first one is always associated with a sensory object and it's a subtle but regular variation between the object existing and not existing. I can compare the rate of these vibrations between objects and for slow vibrations I can even count individual waves such as my finger now vibrating at three times per one heart beat. For more rapid vibrations I sort of "hear" them but I cannot count the individual flashes of existence.
The second type of vibration that I perceive is of perception itself. My attention is constantly resetting itself and returning to the object of concentration. Sometimes it manifests as "bad key-framing". For a split second at a time the experience is frozen in time but still moving in a simple way until another moment of perception replaces it. The simple predictor movement from one "key-frame" to the next doesn't always line up and that's how I notice that my attention is resetting itself few times per second.
I've been doing this for a while now and reached the end that the author calls partial enlightenment. I dislike the term however. It sounds so new agey.
The only thing that has happened is that I now see myself as an inseparable part of the physical universe both on the level of belief and alief. This insight makes life's problems seem less acute as they cannot target a permanent self on the alief-level. I still get sad, feel pain and pleasure but I see them more as fantastic stories about reality made up by my brain than the actual reality of (physical) cause & effect. I can also trust my instincts and subconscious mind more as I no longer fear losing myself when I surrender to pure feelings and motions beyond verbal thought.
This insight is only about a week old and I'm still making sense out of it so I'll end up with a more detailed description of what happened:
I progressed through the first two stages using other meditations but I quit smack in the middle of stage three. Back then I had no idea that feeling awful was a normal part of progress and I spent the next two years depressed in a Dark Night. Please don't do this to yourself! It sucks big time.
Last summer I ran into DavidM's article and started practicing. I worked through the pains of the Dark Night and finally broke out of stage three. However the boredom of early stage four sucked the motivation out of the practice and I took a break until I realized I was slipping back into depression.
So with the support of some friends and strong green tea I started meditating and hunting for the vibrations that permeate all experiences. I finally reached the end of stage four and could see the observer clearly. Like plucking out a flower I picked up the experience of the vibrating observer and realized that it was simply an automatic label the mind puts on thoughts and actions that it finds really important. Having now gained control of the label I could put it anywhere and immediately become it. So I became the universe and the mind was filled with joy.
The habitual mind (that is feeling like my usual self) eventually returned but it now has a dreamlike character and with a little meditation I can attach the feeling of self to basically anything and feel that I become that thing. I can also not attach the self to anything and just let the mind rest in a non-dual awareness.
This only applies because switching to a QM model is computationally prohibitive. QM is generally held to be more true than CM and even if you're trying to optimize for things in terms of CM you're still better off using the QM model as long as you have a good mapping from your QM model to your CM goals.
Humans do indeed find it difficult to think in terms of QM, but this need not be the case for a future AI with access to a quantum computer. If the CM model and the QM model could be run with similar efficiencies then the real issue becomes the mapping from QM model to CM goals. All maps from QM to CM leak in terms of what counts as being located inside the box so the AI might find ways to act outside the box (according to a different mapping). This highlights the point that with computational resources being equal the AI will always prefer the most general available world model for decision making even if its goals are defined in terms a less general model.
I have to point out that the issue with QM to CM mappings is mostly of theoretical interest and in practice it should be possible to define a mapping that safely maximizes the probability of the AI staying in the box while still being able to function optimally. The latter condition is required because a mapping from QM to CM that purely maximizes the probability of the AI staying in box will cause the AI to move in to the middle of the box and cool down.