“QRI Law of Transhumanism”: The overall motivation of humans to solve social and mental problems will remain much higher than the motivation to solve physics problems. The human performance in solving social and mental problems will remain much lower than the performance in solving physics problems. This continues until social and mental problems become physics problems.
– Anders & Maggie
Letter VI: The Evolution of the Phenomenal Self
A follow-up for the more nerdy audience could perhaps be how QRI seeks to resolve the confusion about individualism:
It often turns out that parsimony is a more useful guiding principle in science than naïve realism. This includes naïve realism about what constitutes parsimony. All relevant conditions must be taken into account, and some conditions are unknowns, which blurs the picture. Occam’s razor is powerful but more like a Samurai sword: you need great skill to use it well.
Compare the state-space of consciousness with the state-space of chemistry known to humans: there is biochemistry and there is other chemistry. They manifest quite differently. However, parsimony favors that at the fundamental level of organization things reduce to a small set of rules which are the same for all of chemistry. This is now known to indeed be the case but was not always so. Rather, it tended to be assumed that some extra factor, a “life-force”, had to be involved when it comes to biochemistry.
Biochemistry has been evolutionarily selected for performance on a most formidable problem. That of self-replicating a self-replicator. It takes a large number of steps in the process and high preciseness at each step. Only particular sequences of steps lead to normal cell function, and things are always open to getting corrupted. Take viruses, for instance.
Normal function of a brain is somewhat analogous to normal function of a cell. Evolution has selected for brains which produce the experience of continuity as a unique agent self. This is probably one of the harder tasks that conscious intelligence has solved, corresponding to the advanced parts necessary for reproduction in a cell. It is probably about as unusual in the state-space of consciousness as cellular replication is in the state-space of chemistry. However, the state naïvely feels like it is foundational to everything, which can make you confused when reflecting upon it. It can get even more confusing when you consider the strong possibility that valenced experiences of “good or bad” are much more commonplace in the state-space, perhaps more like transfer of electric charge is commonplace in chemistry.
You can test this by altering (mental) system properties via meditation or psychedelics. Is “individuality” or “valence” more persistent under perturbation? It’s much harder to get rid of valence, and indeed, the highly altered state of a brain on high doses of 5-MeO-DMT gets rid of the agent self altogether but preserves and even enhances valence, interestingly more often in the positive than the negative direction. It’s like jumping from biochemistry to pyrotechnics.
Naïve realism would hold that the sensations of “one-ness” experienced in certain highly altered states of consciousness feel the way they do because they somehow expand to include other entities into a union with yourself. What is likely to really be going on could be the opposite: there is no “self” as a reality fundament but rather a local complex qualia construct that is easy to interfere with. When it (and other detail) goes away there is less mental model complexity left. A reduction in the information diversity of the experience. Take this far enough and you can get states like “X is love” where X could be anything. These can feel as if they reveal hidden truths, for you obviously had not thought that way before, right? “X is love, wow, what a cosmic connection!”
Letter VII: Fractional Crystallization to Enhance Qualia Diversity
Some more chemistry: is there in qualia state-space something analogous to fractional crystallization? When a magma solidifies relatively rapidly, most of the minor elements stay in solid solution within a few major mineral phases. You get a low diversity assemblage. When the magma solidifies slowly it can yield a continuum of various unique phases all the way down to compounds of elements that were only present at ppb levels in the bulk. Crucially, for this to work well, a powerful viscosity reducer is needed. Water happens to fit the bill perfectly.
Consider the computational performance of the process of solidification of a thousand cubic kilometer plutonic magma with and without an added cubic kilometer of water. The one with the added water functions as a dramatically more efficient sorting algorithm for the chemical element constituents than the dry one. The properties of minor minerals can be quite different from those of the major minerals. The spectrum of mineral physical and chemical properties that the magma solidification produces is greatly broadened by adding that small fraction of water. Which nature does on Earth.
It resembles the difference between narrow and broad intelligence. Now, since the general intelligence of humans requires multiple steps at multiple levels, which takes a lot of time, there might need to be some facilitator that plays the role water does in geology. Water tends to remain in liquid form all the way through crystallization, which compensates for the increase in viscosity that takes place on cooling, allowing fractional crystallization to go to completion in certain pegmatites.
It seems that, in the brain, states become conscious once they “crystallize” into what an IIT-based model might describe as feedback loops. (Some physicalist model this as standing waves?). Each state could be viewed as analogous to a crystal belonging to a mineral family and existing somewhere on a composition spectrum. For each to crystallize as fast and distinctly as possible, there should be just the right amount of a water activity equivalent. Too much and things stay liquid, too little and no unique new states appear.
It may perhaps be possible to tease out such “mental water” by analyzing brain scan data and comparing them with element fractionation models from geochemistry?
Eliezer Yudkowsky has pointed out that something that is not very high hanging must have upgraded the human brain so that it became able to make mental models of things no animal would (presumably) even begin to think of. Something where sheer size would not suffice as an explanation. It couldn’t be high hanging since the evolutionary search space available between early hominids and homo sapiens is small in terms of individuals, generations, and genetic variability. Could it be a single factor that does the job as crystallization facilitator to get the brain primed to produce a huge qualia range? For survival, the bulk of mental states would need to remain largely as they are in other animals, but with an added icing on the cake which turned out to confer a decisive strategic advantage.
It should be low hanging for AI developers, too, but in order to find it they may have to analyze models of qualia state-space and not just models of causal chains in network configurations…
Letter VIII: Tacking on the Winds of Valence
We just thought of something on the subjects of group intelligence and mental issues. Consider a possible QRI framing: valence realism is key to understanding all conscious agency. The psyche takes the experienced valence axis to be equal to “the truth” about the objects of attention which appear experientially together with states of valence. Moment to moment.
Realism coupled with parsimony means it is most likely not possible for a psyche to step outside their experience and override this function. (Leaving out the complication of non-conscious processes here for a moment). But of course learning does exist. Things in psyches can be re-trained within bounds which differ from psyche to psyche. New memories form and valence set-points become correspondingly adjusted.
Naïvely it can be believed that it is possible to go against negative valence. If you muster enough willpower, or some such. Like a sailboat moving against the wind by using an engine. But what if it’s a system which has to use the wind for everything? With tacking, you can use the wind to move against the wind. It’s more advanced, and only experienced sailors manage to do it optimally. Advanced psyches can couple expectations (strategic predictive modeling) with a high valence associated with the appropriate objects that correlate with strategic goals. If strong enough, such valence gives a net positive sum when coupled with unpleasant things which need to be “overcome” to reach strategic goals.
You can “tack” in mental decision space. The expert psycho-mariner makes mental models of how the combinatorics of fractal valence plays out it in their own psyche and in others. Intra- and inter-domain valence summation modeling. Not quite there yet but the QRI is the group taking a systematic approach to it. We realize that’s what social superintelligences should converge towards. Experiential wellbeing and intelligence can be made to work perfectly in tandem for, in principle, arbitrarily large groups.
It is possible to make a model of negative valence states and render the model to appear in positive valence “lighting”. Sadism is possible, and self-destructive logic is possible. “I deserve to suffer so it is good that I suffer”. The valence is mixed but as long as the weighted sum is positive, agency moves in the destructive direction in these cases. Dysfunction can be complicated.
But on the bright side, a formalism that captures the valence summation well enough should be an excellent basis for ethics and for optimizing intelligences for both agency and wellbeing. This extends to group intelligences. The weight carried by various instantiations of positive and negative valence is then accessible for modeling and it is no longer necessary to consider it a moral imperative to want to destroy everything just to be on the safe side against any risk of negative experience taking place somewhere.
At this early stage we are however faced with the problem of how influential premature conclusions of this type can be, and how much is too much. Certain areas in philosophy and ideology are, to most people, more immediately rewarding than science and engineering, and cheaper, too. But more gets done by a group of scientists who are philosophically inspired than by a group of philosophers who are scientifically inspired.
Could this be in the ballpark-ish?
Stay safe and symmetric!
– Maggie & Anders