Missing Pieces Of My Mind
Smart people really enjoy advertising how bright they are by showing off how much they know. This is a particularly irritating conceit of scientists who are supposed to be in the business of studying the UNKNOWN not validating all the stuff they already believe to be true that probably isn’t. I like writing articles about human ignorance because it’s a lot easier to find interesting scientific problems to study when you aren’t utterly delusional about what you think is true. Understanding the boundaries of human knowledge is also a really powerful tool for separating valuable “information” and insight from BS. I’ve written a lot of articles about things that the human race has no idea how to compute yet that people constantly claim we can. Artificial Intelligence being a particular favorite theme in this genre of ignorance exploration.
“Over time I think we will probably see a closer merger of biological intelligence and digital intelligence… It’s mostly about the bandwidth, the speed of the connection between your brain and the digital version of yourself, particularly output,” Musk concluded. “Some high bandwidth interface to the brain will be something that helps achieve a symbiosis between human and machine intelligence and maybe solves the control problem and the usefulness problem.”
Can I call Elon Musk an idiot for saying this stuff or is promoting AI hysteria some sort of clever PR gimmick that keeps Tesla’s stock price elevated despite never turning a profit? I’m proud to say… “I don’t know”. I have no idea why famous “geniuses” say stupid stuff like this constantly in the press.
In previous blogs I’ve written a lot about why we’re not even remotely close to simulating any kind of human like “artificial intelligence” with a computer. To summarize all of my previous AI blogs in a few bullet points, they can be grossly categorized as follows;
- Biological systems don’t compute ANYTHING LIKE our computers do, they’re not even remotely related ideas in computing
- Our computers can’t even remotely begin to simulate complex biological processes
- We don’t know how to simulate complex biological processes even if we had computers that could do it
- Neural networks are clearly nothing like credible models for a human mind
- Neural networks don’t self-program or create
- Neural networks are marionettes that scientists use to delude themselves into thinking they’re mastering intelligence
- We have no idea what’s going on inside the cells that MAKE our brains
- 90% of the human brain is composed of glial cells which we believe do nothing interesting other than assemble our minds from scratch and control our thoughts in ways we don’t understand thus hardly meriting closer examination
- Evolution is not random, living systems appear to selectively mutate themselves on purpose, i.e. they deliberately self-program at the genetic level
You get the idea. I get my cred by clearly illustrating how stunningly ignorant everybody else is.
[…] ἔοικα γοῦν τούτου γε σμικρῷ τινι αὐτῷ τούτῳ σοφώτερος εἶναι, ὅτι ἃ μὴ οἶδα οὐδὲ οἴομαι εἰδέναι.
[…] I seem, then, in just this little thing to be wiser than this man at any rate, that what I do not know I do not think I know either.
I find that one of the most fascinating things about being human is attempting to use my mind to analyse itself! For example, if it’s really a neural network we should be able to imagine a mental test that supports or invalidates this hypothesis. For example it should be impossible to think about things for which there are no connections in our mind. Quick, name two completely unrelated things! Suppose you pick the words “Duck” and “En Passant”. Go ahead make a connection between them! If you can’t think of a relationship between these two words… how did you think of them in a pair? Did some sort of mental anti-association random number generator in your mind pick them out? Do neural networks do that?
In 1986 the legendary MIT father of AI, Marvin Minsky published his seminal book “Society of the mind“.
It was tremendously influential in shaping my ideas on AI at the time. One of Minsky’s valuable observations is that the mind seems to be composed of many little conscious agents all competing for control of the marionette strings that control our bodies. Each agent is obsessed with a different problem, some want food, some want sleep, others want to watch TV. These agents are constantly competing with one another for control of our behavior. As hunger, thirst or boredom grows the associated agents gain influence and are able to suppress the chorus of voices to take control. Collectively we associate this sequence of influences with a single conscious experience. It’s an interesting theory but is there any evidence that this is really the way the mind works?
“Individuals with schizophrenia may experience hallucinations (most reported are hearing voices), delusions (often bizarre or persecutory in nature), and disorganized thinking and speech. The last may range from loss of train of thought, to sentences only loosely connected in meaning, to speech that is not understandable known as word salad. Social withdrawal, sloppiness of dress and hygiene, and loss of motivation and judgment are all common in schizophrenia.
Distortions of self-experience such as feeling as if one’s thoughts or feelings are not really one’s own to believing thoughts are being inserted into one’s mind, sometimes termed passivity phenomena, are also common. There is often an observable pattern of emotional difficulty, for example lack of responsiveness. Impairment in social cognition is associated with schizophrenia, as are symptoms of paranoia. Social isolation commonly occurs. Difficulties in working and long-term memory, attention, executive functioning, and speed of processing also commonly occur.“
Kind of sounds like the sort of experience one might have if your mind where made of a chorus of competing conscious agents that couldn’t take turns pulling the marionette strings, doesn’t it? Like they couldn’t suppress one another properly? Schizophrenia is also a common and seemingly easily triggered disease, not a rare genetic fluke exclusive to a few unfortunate mutants.
“Problems with certain naturally occurring brain chemicals, including neurotransmitters called dopamine and glutamate, may contribute to schizophrenia.”
So in a Minsky AI, the brain is not a highly coordinated cohesive neural network that processes and adapts to new input in a highly structured regulated way like a Bolshoi Ballet, it’s more of a bar room brawl for control of the big-screen remote control. More importantly, Minsky imbues every sub-component of the mind with a little consciousness, a little free-will so to speak. The challenge with Minsky’s idea is that it still leaves the source of the magical property of consciousness unexplained. What’s missing from this model?
- A root source of consciousness
- A mechanism for building brains and a nervous system out of ignorant cells without a brain or nervous system to coordinate the process
- Some other stuff I can’t think about yet
Let’s tackle the subject of how an intelligent organ that “thinks” electrically via a network of neurons gets constructed by a swarm of unnetworked cells.
First interesting observation? The NEURONS that actually construct the brain don’t communicate with one another electrically. These cells also make up 90% of our adult brains. Do you suppose they’re doing anything important?
“Brain Cells that Communicate without Electricity: Calcium Waves in Glia. Glia are brain cells that cannot generate electrical impulses. … Probing the brain with electrodes, the way neuroscientists do to understand neuronal communication, is useless to intercept glial communications.”
These are the cells that appear to “assemble” the mind by extending fibers that guide neurons into position during fetal development.
“Although it has become increasingly evident that glial cells
play an important role in neurogenesis from the earliest formative
stages, many issues, including details of glial cell
lineage, the nature of their interaction with neurons, and their
function during various phases of ontogenetic development need
to be clarified.“
In other words, we have no idea how the brain is constructed by neurons that aren’t wired up to communicate electrically. If glial cells are coordinating an effort to construct a multi-billion cell mind, they’re doing it by communicating via surface-to-surface contact with one another and/or via slow long range messaging via the circulatory system. If these modes of communication are sufficient to construct a mind, then they are extremely powerful and unlikely to cease playing a role once the mind is assembled. We already know that our DNA does not contain a “blue print” for our bodies, just rules that cells follow that result in creating us. We also know that even identical twins do not have identical cellular physiology.
“You would expect identical twins to have the same fingerprints as they are monozygotic, which means that they develop when a single fertilized egg splits in two, leading to two embryos. As they both came from the combination of the same egg and sperm, these twins have virtually indistinguishable DNA.
So why aren’t their fingerprints identical from birth? Fingerprints are not entirely a genetic characteristic. They are a part of a ‘phenotype’ which means they are determined by the interaction of an individual’s genes and the intrauterine environment (differing hormonal levels, nutrition, blood pressure, position in the womb and the growth rate of the fingers at the end of the first trimester).”
In other words, cells appear to have a lot of “creative” latitude in how they build our bodies successfully. If you really wanted to understand how to create an “artificial intelligence” isn’t this property of cells what you should be simulating? In theory you only have to be able to simulate a SINGLE fertilized egg cell to get it right.
Suppose for the sake of discussion that you felt really strongly that you needed a brain to correctly construct another working brain… then whose brain is available for that job? Mom’s brain!
What are babies cells doing migrating to mom’s tissues and brain? One possibility is that they are there for training before migrating back to the fetus and some of them get marooned and incorporated into mom. This creates an unfortunate chicken-and-egg paradox for simulating mammalian brain development… literally… How would we test the idea that humans need their mothers brains to develop their own brains successfully? We would need to try growing an infant outside a womb such as was accomplished recently with this sus-vede lamb experiment in which scientists grew a baby sheep outside the womb in a plastic bag.
Unfortunately if you read the article closely you’ll see that they didn’t transfer the fetus to the bag until after 100 days of fetal development… its brain was already grown. Of course you’d never read a scientific article about how they FAILED to grow a lamb in an artificial womb, which suggests that maybe they tried it with even more premature fetuses and it failed. So we can’t discount the possibility that a working adult brain plays an essential role in the correct development of a working mammalian mind.
But we digress… let’s get back to the role of glial cells in brain activity.
“Calcium is also released randomly and without stimulation from astrocytes’ internal stores in small bursts called ‘puffs.’ These random puffs can lead to waves. It is possible that the seemingly random thoughts during dreams and sensory deprivation experience could be calcium puffs becoming waves in our astrocytes. Basically, it is obvious that astrocytes are involved in brain processing in the cortex, but the main questions are, do our thoughts and imagination stem from astrocytes working together with neurons, or are our thoughts and imagination solely the domain of astrocytes? Maybe the role of neurons is to support astrocytes.”
This guy is taking all of the fun out of being a skeptic, he’s got a PhD and agrees with my crazy theories. Here he’s proposing a physical mechanism by which “creativity” emerges from glial cells and mediates our thinking. Note the presence of a key word that scientists like to throw around casually that basically translates to “magic” or “I have no idea”. The word is RANDOM. Do we believe that glial cell “puffs” are random? What he is really saying is that there is some genetic mechanism inside the glial cell that we don’t understand that systematically generates these “puffs” in order for our brains to work correctly! Recall my earlier treatise on the essential role of carefully chosen RANDOM weights to making neural networks work correctly? Our artificial neural networks only set these weights as initial conditions… they don’t keep changing randomly which is one reason that our human made neural networks don’t make abrupt leaps in insight. If they were really random they might even make the neural network perform worse. Hence my contention that intelligence and consciousness as we experience it arises at the genetic level from our cells and has extremely primal origins. I’m not saying that this is the exact mechanism by which it occurs, just that this approach to thinking about the problem produces much more satisfying answers about how intelligence works than any of the prevalent views about simulating intelligence with a computer.
This theory is also consistent with a much more esoteric book on fetal development that I read some years ago in which a really anal-retentive researcher hand mapped the first few hundred cell divisions of dozens of species with a microscope and drew detailed diagrams of every early cell division stage and early organ formation. He also mapped the genetic signals issued by each cell as they specialized and documented how interfering “waves” of hormones released by these cells interacted with one another to form discreet organs. *I confess that I used to use Mathematica to build cellular automata simulations of sea shell development back in the 1990’s influenced by the brilliant work of Heinz Meinhardt and his colleagues.
Simulating this kind of cellular signaling in liquid environments is quite complicated but sea shells are conveniently very rigid structures and as you can see the results are very convincing. The activation/suppression model of inter-cellular communication appears to be akin to a slow motion version of three dimensional holographic waves propagating and interfering with one another across touching cell walls with each cell modifying the signal slightly at each stage. This suggests a mode of computation that is similar to a neural network but vastly more powerful and primal. This is perhaps the underlying mechanism by which cells “think” and “coordinate” that gave rise to neural networks that function on a similar principle except later evolved to use electricity to compute faster and over longer ranges. In other words a bunch of cells can “think” even before they’ve formed a brain.
There’s much more I want to discuss on the subject but this blog has already covered a lot of ground. What I want to cover next is the specific genetic machinery INSIDE cells that do the thinking and where the mysterious and apparently essential “RANDOM” component of thought emerges from.