Thanks as always to you as always, readers, again stopping by the @Lamb_OS Substack. It’s truly always a pleasure to have you stop by.
By now, if you don’t yet know me, then it’s unlikely another rant about it will help (hasn’t yet, it seems). So let me begin and end by introducing myself as Dr. William A. Lambos, neuroscientist, data scientist, and licensed neuropsychologist. I write a lot about AI, but not exclusively. See my previous screeds on this Substack to learn more.
Please Subscribe today if you have not already yet done so. As I like to dissemble, I’m closer than ever to getting that private jet, and every new subscriber earns me…$0.00. Yeah, so the “private jet” thing is – and will sadly forever remain – an utter (if enjoyable) delusion. After all, this Substack is still free! So, please subscribe to support my…self-concept? You decide – but please, only after you subscribe.
§
OK, on the “Molecules of More”.
The graphic above is a slide from a talk I’ve been giving for the last six months or so. The presentation is named “A.I. Is Neither”, an admittedly snarky opinion embedded in a PowerPoint title.
But I believe the opinion itself – snarky or otherwise – that AI systems have yet to show anything remotely like human intelligence, is more than defensible. It is in fact more of a statement on the nature of the AI beast. The stronger assertion, and the one I will maintain until evidence shows otherwise, is that AI inherently andnecessarily lacks the substrates of intelligence itself. If you insist, I can be more definitive: AI lacks the ability to do (or act in) nearly every manner that humans mean and imply when we use the term “intelligence”, at least when we do so without the reference to computational automata. Note that the last criterion is telling in and of itself. More on that near the end – so read on!
As to what these “substrates” are, I offer them (yet) again, briefly, here:
Intelligence implies functional and adaptive behavior, which itself can always be shown to be motivated to address a need or goal.
The ability of an entity to engage in the circle of motivated behavior and goal attainment is summarized by the term “Agency.”
Therefore, the systemic intertwining of goal attainment and adaptive behavior within a frame of reference – that is, intelligence – is made possible only because entities capable of intelligence possess “agency.”
The premises and conclusion above are of valid logical form. If the premises are true, so must the conclusion be. However, it is not my intent to continue arguing in favor of the truth of the premises. I’ve done that in my last four posts. Read them now if that’s what you are looking for.
Rather, I want to focus on the nature of this agency thing. If I can reasonably make the case that agency itselfrequires a substrate not present in computational systems, and currently impossible to implement in them, then at least we can return to arguing over whether the syllogism’s premises are true. Which would be, as we said in the third grade, “one giant step forward.”
So, where does the agency (that I argue) is necessary for human-like intelligence come from?
Let’s return the main part of the slide above, made larger below. It is reproduced from Leuchter AF, Hunter AM, Krantz DE, Cook IA. Rhythms and blues: modulation of oscillatory synchrony and the mechanism of action of antidepressant treatments. Ann N Y Acad Sci. 2015 May;1344(1):78-91., 2015. The graphic proposes to elucidate 10 levels at which causal factors influence, bidirectionally, one another. The gamut runs from the genetic constituents of the cells in a living organism (at the bottom), and the resulting observed expression of these proteins at the so-called phenotype of the intact organism at the top – in this case, a cut-out that represents a human being.
OK, great! So what?
Let me be the first to admit that I looked at this graphic for many years before I became aware of any connection to differences between biological and computational systems inferred by this representation. But when giving a lecture that included this graphic 2017, I was struck with a realization on which I now base the entirety of my current views of so-called A.I.
The most condensed version of said realization is:
This is a hierarchical structure of the two-way causality associated with every level of biological attributes above the level of physics. It is immediately reminiscent of Maslow’s “Hierarchy of Needs.”
Is it a coincidence that Maslow’s Hierarchy also starts by placing biological needs at the bottom of his pyramid, and puts loftier needs such as self-concept and career goals, well above them? I do not think it is…
Where then, ultimately, do all the needs of living creatures originate?
FROM THE DNA INSIDE OUR CELLS!
Which immediately implies (but does not prove) that without the influence of DNA, there is no other readily available explanation underlying goal-driven behavior. And by “readily available”, I mean, show me just one – any single one at all! You can’t. Neither can I. And I’ve been looking for one, at least intuitively, for more than 50 years!
Until such time as someone can demonstrate otherwise, DNA is, then, the very basis of all organized and goal seeking behavior. The basis which makes possible the propensity for living things to compete for resources, and (sorry to Karl Marx), to impose their will upon the world.
Also often overlooked here is that the competition for the resources that enable survival happens at the level of every cell within living organisms. Ever stuck yourself with a fork (I mean, literally!)? I hope not, but every time when I did, I noticed that some areas under the bandage healed faster than did others (yes, as embarrassing as it is, I did it more than once. This may be correlated with the suggestion offered by my Troop leader to give more thought to “re-upping” to camp out with the other scouts in my pack).
Why should healing not be uniform across areas? The answer is, when adjacent or nearby cells in multicellular organisms suffer tissue damage (even in the same finger), these cells must compete for the bodily resources needed for repair.Only the strong cells get the goods- those that “shout” the loudest, in terms of excreting chemical signals (i.e., cytokines, in vertebrates) to attract the increased blood flow to needed for delivery of the oxygen, glucose, immune cells, lipids, proteins, and the many other “resources” needed for reconstruction.
Go back far enough in the evolutionary record, and one will see the same trajectory between whole organisms and, of course, between species: the struggle to edge out the competition. This may involve anything from to development or improvement of sensory-motor mechanisms for learning where to go and what to do to meet one’s needs (and, more importantly, where and what to avoid). Follow individuals and species through further natural selection and watch – in evolutionary time – the emergence of memory, of the sensitivity to consequences, of the ability to recognize generalities from individual exemplars, and for everything else necessary to survive, reproduce, and, one would hope, prosper, from doing “commerce with the environment.” Interested readers may wish to see the writings of Edmund Tolan and his “purposive behaviorism” for wonderful examples of how adaptive behavior becomes guided (i.e. “intelligent”) on this basis, without any reference to the internal mechanisms imbued in the organism (and which, in truth, likely added to the myth of artificial intelligence).
Finally, it should not be a stretch to see that virtually everything else we imply when we use the term “intelligence” are subsumed by the DNA rising from within. All the way from the skeletal reflex that causes us to withdraw our (my?) fingers from the tines at the end of the fork, to functions of the brain’s frontal lobes that underlie “executive functioning”, including the abilities to navigate a social and technological world made exponentially more complex by…us. Finally, and most importantly, it is the expression of DNA, and, I believe, this expression alone, which allows human beings to build empires, unrivaled among other species of living creatures. From dramatically different points of view, both the neuroscientist and Numenta CEO Jeff Hawkins’s and the Oxford-trained (now at Hebrew University of Jerusalem) Yuval Noah Harari have recognized that such achievements owe more than anything else the ability to reasonably predict the future. Hawkins notices this when discussing how the neocortex continually compares expected to actual experiences and outcomes detected by the brain (within individuals); Hurari focuses, for his part, on describing how we predict, prepare for, and attempt to modify a shared future through the telling of “stories” when groups of us speak among amongst one another.
To me, it is fascinating that both continuous neocortical prediction (believed to be highly correlated with the activation of the brain’s salience network) and the campfire stories love to we scare ourselves with, have the same purpose: to predict the future, and thereby better prepare for it. These abilities, both of which depend entirely upon the existence of DNA, are the basis of all we mean when we use the term intelligence, at least if we are honest with ourselves (questioning the meaning of intelligence when we compare ourselves to automata seems more likely to reflect hubris than otherwise). Hurari himself notes how ironic it is that many of us are focused upon how we will survive the presumed “AI Apocalypse,” instead of worrying about whether the possibility exists for any such outcome.
Is it really such an impossible leap to make to grasp that DNA not only does, but – thus far – must underlie agency, and therefore, intelligence? That not only do agency and intelligence owe to the connection between themselves and DNA, but to their dependency – at the cellular, organism, troupe, and global levels?
Most importantly, should you, the reader, think it is so far-fetched to suppose that without DNA’s goal seeking “agents” burrowed within the deepest levels of living tissue, analogs to true intelligence can be expected to arise from computational architectures based on silicon?
Honestly, I hope not!
Thanks as always for reading and see you next week!
Bill
“Be kind, for everyone you meet is engaged in an enormous struggle."
— Philo of Alexandria
William A. Lambos, Ph.D.,
Licensed Psychologist, BCN
Neuroscientist, Data Scientist
CNS Computational Neuroscience
walambos@mac.com
813.235.4270 Office
888.503.3166 Fax
I agree that nothing in current "AI" technology has a chance of leading to anything approaching true intelligence or agency.
But it's a legitimate question whether we could (eventually) build a system in which all aspects of a man-made system, from virtual DNA on up, mimic in meaningful ways the complex functions of physical biological systems. And if the answer is yes, could we then create something that has virtual agency? (I'll leave aside the question of whether or not we *should*, since even if it's clear we shouldn't, some rationalizing person somewhere will attempt it if the technology becomes available).
I think the theoretical answer is probably yes. But I think most people underestimate the amount of resources (memory, processing power, etc.) required for such virtual biology by many orders of magnitude. Enough orders of magnitude, in fact, that even if Moore's Law continues unabated forever (which is doubtful) the chance of us achieving anything like this within several centuries seems small.
One of the reasons for this is that, contrary to what many digital bloviators would have you believe, neither our brains nor any other part of us are digital. And no part of us would lend itself well to digitization without significant loss of fidelity (which in this case would mean intelligence). IOW, we are analog through and through, and as any audiophile can tell you, analog inherently captures volumes more data than digital. So, the amount of information packed into a single cell of our brains and bodies is staggering, even by modern megadata standards.