Elon Musk is Wrong: Why AI is a Red Herring

I want to make it clear that this is not meant to insult or degrade Elon Musk or his followers, rather to give them a viewpoint they may not have considered so they can re-evaluate their views and see if they come to a different understanding. I know Elon Musk has been a controversial figure. And zoomers, which I assume make up a massive amount of his fanbase, have a reputation for being stupid. But maybe zoomers actually are on to something with their Dogecoin schemes.

The argument against AI reaching human consciousness in the near future is simple and has been stated before: the human brain's level of complexity is too high to be replicated with our current computer tech, and it will take an excruciatingly long amount of time for anybody who wants sentient AI to get it. In fact, they'll be long dead by the time we get it. But despite the simple truth of this argument, many people cite accelerationist logic and the technological singularity hypothesis as legitimate possibilities that could allow for advanced AI arriving in the world soon, so the argument has to become more specific to demonstrate why AI cannot catch up in such a short spell of time.

The way Ken Wilber articulated why human consciousness cannot be downloaded to a machine in the near future was essentially just the complexity argument, although it was stated perfectly in a poetic way with the fact that uncountable billions of neurons in a system would be difficult to replicate when he alleges there are more neural connections in the neocortex alone than there are stars in the "known universe", especially using modern computer science techniques that don't seem to be advancing as fast as accelerationists believe. "Oh, wow, it got smaller." The amount of transistors and such will take time to catch up to neuron counts, and even so, neurons are more versatile in terms of logic whereas transistors are usually more one-dimensional in function design.

Machines are also susceptible to the flaws of humans and limits of what technological resources we have access to. For example, think of the way a human being thinks about people, locations, or other objects they have seen before. Their eyes intake the photons with a very specific acuity. The specific image "imprints" the brain, with the parts of the brain stimulated associating mostly with the memories of that thing. This can generate linguistic logical information that is associated with the visual associations and can be used to conceptualize about the object being seen and therefore make decisions based on which of the associated hypotheticals associates most with dopamine release, which causes the "association" between the mind's decision and the body's action. It's all connections based on how neurons structure after billions of years of effective evolution, whereas humans are having a harder time developing this because humans are slow and have to do things like interrupt their work in order to eat, sleep, etc in a way where simply designing all that in addition to the countless other functions the brain has would make this task take ridiculously longer than the near future. If it to happen at all, the most efficient method will be humans just reflecting exactly the sorts of systems seen in the brain, without knowing what actually's going on (the same way the Youtube algorithm is a mystery even to Youtube or how a chess engine designer does not have access to all the decisions of a chess engine). The problem here is that neuroscience and the mapping of the brain's mechanisms themselves are very slow.

It's like a metaphor I used in another blog post, the one about intelligence.

The basic theoretical mechanics of an engine are just energy pushing an object, but there are more complicated structural variances like multiple pathways to maximize energetic efficiency and coolant mechanisms to prevent overheating and injector mechanisms in spaceships for everything to be done in proper ratios and angles, etc.

I also question why someone would want to live among these hypothetical machines anyway. Sure, perhaps machines can be more efficient at solving certain problems, but highly intelligent people can already solve the problems that actually matter. A computer can solve tiresome mathematics like the square root of (19191919191+737373737337)/2+9(282828) in no time, whereas even someone like Einstein or anyone short of some walking calculator mental math prodigy would not compare. But when it comes to the stuff that actually counts, AI would say the same things as any reasonable person. If running an accurate simulation, the machine learning would see the way to end world poverty and war and famine and pestilence is for all the humans to converge together slowly over time into oneness. If that search result was "banned" from the AI questioning for "practical" capitalistic purposes, the AI could find other ways to sneak around the stupid owners. But perhaps the owners could game the AI to not be targets, the same way Timothy Leary was able to be sent to minimum security prison and escape because he had helped design the prison psych eval at Harvard and so knew which specific answers caused him to be considered the most stable and low risk. Genetically perfected humans could probably serve the same purpose as AI, and they would be easier to manufacture, even if I have a sense there's something very, very wrong with how that could be used: if all differences and "realness" between humans were erased. This is basic instinct, but Plato stated it in the philosophical language through saying we should not believe "the world was made in the likeness of any Idea that is merely partial" because "nothing incomplete is beautiful. We must suppose rather that it is the perfect image of the whole of which all animals--both individuals and species--are parts."

Fear-mongering about AI does not make sense even if it arrives. Just follow basic safety precautions like don't give the AI a powerful physical body, don't give it access to nukes, give it a pre-built cost function for the decision of attempting to destroy the human race, monitor stuff it decides to manufacture, etc.

But all this effort could be futile anyway, when it comes to replicating advanced consciousness.

Wilber makes the point that if you download information from the human brain into a hard drive in a way that is dissociated and altered from how it is within the nature of the brain, the consciousness will be dissociated and altered from its original nature as well. This principle is related to how Wilber and the philosophers that influenced him consider holonic understanding to be important. You cannot replicate a molecule without proper atoms, and you cannot truly replicate a human consciousness without the natural components. In Sex, Ecology, Spirituality: The Spirit of Evolution, Wilber lists examples of holonic systems as part of his integrative AQAL model of philosophical cosmology where more complex systems arise from simpler systems or structures. For example, the material in a galaxy can coalesce into planet (the opposite of entropy), planets can foster a "Gaia system" or a biosphere, heterotrophic ecosystems can arise after that, social societies of sapient beings with division of labor can arise after that, families and tribes can form, early states can form, empires can form, and eventually planetary systems with more social and physical complexity can form. It is this running up causative pattern of simplicity to complexity that correlates with how you cannot have an exact whole without its exact parts (and additionally it is easier for a complex thing to come from a slightly less complex thing, rather than produce intense complexity from very mere simplicity, which would be like a massive jump from human-made computers to sentient minds).

I will explain this principle and use a visual formulation of how AI designers could work around it.