AI and Quantum Technologies: Sharing a (Mis)understanding

In recent years, media attention toward the different capabilities (and threats) of AI has been unceasing. That commensurates with the rise of many talking-heads on the topic with hot-takes on the issue, from optimism to warnings for the end of mankind (just browse YouTube). However, that fixation is also moving to what had initially been a slow-simmering, but now, escalating interest, in quantum technologies. Depending on who you ask, present-day AI is the culmination of a gradual technologization process of what were, for the longest time, thought-provoking concepts embodied by symbolic logic that resides at a corner of theoretical computer science, while also moonlighting as a favorite thought-experimental object in the philosophy of mind/cognitive science.

However, improvements to computing processing power and storage capacity (think cloud computing), research development in computational linguistics and natural language processing, algorithms for processing large language models that are more than just texts (the consequence of data accumulated over the course of generations of social media apps), as well as the capacity of neural networks to engage in multi-dimensional informational processing and simple decision-making based on past learning, mean that the theoretical concepts residing primarily in thought experiments could now turn the entire internet-connected world into its field lab.

The aforementioned areas of research, that had once existed separately, have since converged into AI research by way of machine learning. If you would like a deeper look into the reality and hype around AI, I suggest looking here (for an industry perspective) and here (for an academic perspective). While present policy interests in AI and quantum technology — or technologies, the oft-used plural to denote multiplicity in points of emergence or potential, though I suspect it is more of a rhetorical strategy — appear to take these two as separate technological species, their shared and overlapping technical genealogies suggest otherwise.

First, I start with a historical perspective: the ideas that gave root to AI took place around the same time as the ideas that gave birth to what we now refer to as quantum technologies. It began when theoretical and philosophical contemplation into the potential of physical sciences, stemming from several revolutions and evolutions that changed both experts’ and the public’s view of science, gave way to a mechanical revolution that could automate much more of labor, thereby serving the needs of the first Industrial revolution in areas such as mechanical capabilities and computational efficiencies.

Then, the late 19th century and early 20th century saw certain developments in electricity, magnetism, heat, and optics; all of these were further aided by developments in the field of abstract algebra and geometrical algebra; bringing about what we now know as theories of relativity (both special and general) and quantum mechanics. In the case of the latter, some contemporary physicists have argued that the quantum information approach is the best interpretative representation, with some concluding that our actual world is really mostly quantum reality in the same way that it had been suggested that our universe comprises largely of unseen dark matter and energy.

However, to take a step back to the period between the first two world wars, the pioneers of quantum mechanics (I calculate that there are about 2.5 generations, with the .5 representing a sort of overlap between the third and second generation) were not confining their interest to esoteric debates over axioms, theorems, and interpretations, but were also curious over how some of the thinking that they had developed through deep work in physics could be applicable to other areas such as biology, the mind, and the computer.

One of them, Erwin Schrödinger, had written a book where he extended his interest into the biological sciences (genetics more specifically) from the perspective of biophysics, while another one, John von Neumann, had decided that biophysics (and neuroscience) could be one approach for building more powerful computational systems. The work of these two men, and their peers, had contributed to the ‘Atomic’ age and the technologization of thermonuclear fission.

But less talked about (outside specialized communities) was the interest that these two men had in consciousness, intelligence, and operation research (the biological and the computational). By the time the Second World War rolled around, the second industrial revolution had already been around for quite awhile (pretty much since the end of the late 19th century) and the third industrial revolution was just right around the corner. By then, there was an urgent need to start operationalizing the potential of even the most obtuse science to win the war and gain national/international prestige. But most importantly, there is a need to create a system for the scaling of global operations, be it warfare or communications, and cybernetics was the promising route. And cybernetics was where quantum physics and AI intersected.

In a 1989 book (a revised edition was issued in 2016), mathematical physicist Roger Penrose (also of the 2022 Nobel Prize in Physics fame), had published on how quantum mechanics could be the key to unlocking the operating mechanisms of human consciousness. He was not the only person to argue this, since human consciousness has always been a part of the narratives around quantum mechanical interpretations. He claimed that we did not quite grasp artificial intelligence because of how we had associated its developments with what we understand about classical computers. If the premise of artificial intelligence had been approximating both the mechanisms of the human brain and sentience, could we consider all these emphases on algorithmic learning strategies and highly-competent big-data model management as a representation of AI; or do we need to reconfigure the original meaning of this term to better represent what AI has come to be in the present setting, one primarily driven by different methods of training and learning that are highly algorithmic?

Nestled in Penrose’s criticism is the possible notion that a true AI system is likely to be non-deterministic and probabilistic, in the same way that quantum mechanics is, because, if we were to associate AI with consciousness studies, and if consciousness is in any way entailed by quantum mechanics, then the ontologies underpinning AI are naturally probabilistic.

This leads me to the second reason, which is whether all the hype regarding the existence and real-world deployment of quantum technologies and AI has caused us to neglect the distinction between their conceptual vis-à-vis technological realities? At this time, I will not venture into a discussion of whether present approaches to responsible AI and responsible quantum are premised on a sound or misguided understanding in thinking about the relationship between harm and good, or whether it is proper to treat them as different even if their social histories had diverged considerably. Rather, I would like to focus my second reason on a shared constraint that had also shaped their technologization process, and I will divide my analysis into two parts. The first is related to theoretical thinking versus practical realization; and the second is concerned with problem-solving.

In the first instance of theoretical thinking vis-à-vis practical realization, we have to consider the possibility of realizing convoluted mental constructions (and theoretical interpretations) versus scaling up a clear idea with complex parts that merely require rounds (and years) of iterations. Even as far back as 1929 (a historic period for the development of quantum physics), German philosopher of physics Hans Reichenbach had pointed out that technology viewed science merely as a resource, and any scientific investigations would have to be driven by an instrumental need to solve the problem of real-world applications. When dealing with something as indeterministic as quantum physics, technologists would prefer to focus on the measurably tangible that they could then embody in the form of scalable hardware — meaning that they should focus on regions that could be determined, even if its universe is probabilistic.

Since a physical quantum computer is sometime from being realized, what we have in operation are systems that could emulate a quantum computer, in the same way that we could emulate older operating systems in our present one to, for instance, play vintage desktop games. At the same time, as I will show in articles to come, certain forms of quantum-thinking are already entangled within devices seemingly operating only at the scale of classical physics. As noted elsewhere, even when quantum computers are actualized in everyday use, classical computers will likely to exist for some time, though possibly reconfigured through the awakening of some of its presently latent features, either in terms of its circuitry design, processor technologies, or storage capacity management, among others. Moreover, if a quantum computer is akin to a jet-engine, we may ask if it is necessary for a jet-engine to be deployed if we are just going to go to a grocery store 10 minutes away from our home.

In the second instance, we have to consider for whom and to what end, is a quantum computer, and its accompaniment technologies, being developed and tested. At this time of writing, there is still research gap between fundamental research into quantum computing and its deployment capacity within an industrial setting. However, there is recognition of the growing import of quantum physics in the life-sciences through a growing community of interdisciplinary researchers who engage in quantum biological research. Quantum biology, at present, attempts to reconcile the present limitations in the at-scale exploration of biological systems by penetrating and operating into territories unprecedented, i.e. at a level more fine-grained than a molecular structure.

Obviously, any technologies within the life sciences that require greater precision and refinement in techniques and instrumentation will benefit tremendously from the instrumental translation of quantum physics, such as drug delivery and chemical transport. My own research into nuclear technology had included looking at an irradiation technology use case of mutagenesis (genetic mutation) in agronomical practices of developing economies, showing that such applications are neither speculative nor overtly locked within the high-tech. However, the process of irradiation to generate mutagenesis in this instance is almost like a blackbox, where the attainment of the end-result comes through a process of trial and error. In their book, biologist McFadden and physicist al-Khalili had suggested that a better understanding of genetic mutation would require knowledge of what happens at the subatomic, quantum scale. In their 2021 report, McKinsey has suggested that quantum computing would have impact on both the automative and financial sectors, through a process of optimization, at the algorithmic and computational level (one akin to the travelling salesman problem where you solve networked decision-making. This problem obviously had their precedence in AI, since solving the travelling salesman problem is about process optimization. Moreover, the two instances discussed in relation to quantum technologies are also applicable to AI, because our present view of AI that appears deterministic, despite the seeming range of uncertainty with regard to what we consider to be its impact on our future, is driven by how it has been instrumentalized to solve problems, rather than what had motivated its original form of inquiry. In this respect, AI and quantum technologies are similar.

What makes AI and quantum technologies different in the viewpoint of policy-makers is because their operational starting point had always been the technological rather than the ontological, and both of these technologies exist in separate spheres. While this itself is not a wrong approach, one may ask if such policies are sufficiently accounting for the gap between the theoretical thinking that had given birth to these technologies (signaling that what we could operationalize today is highly constrained) and one predicated on problem-solving (which had also been part of the AI for Social Good’s mission). The present focus on ethical, legal, and policy implications of both these technologies had largely been driven by a kind of value-belief-norm (VBN) shaped by a particular version of instrumental rationality, with a multi-linear (rather than chaotically complex) perspective of how an assumed/expected end determines the process of getting to that end.

There is still much about the present technological world that are under-explored, that, if we were to approach using narrative design and forking stories, could unwrap some delicious wonders. Along the way, I may also comment on various strategic directions, be it from a policy or industry (business) standpoint, and move between various possible realities, while returning, again and again, to questions of justice, ethics, and distributed responsibilities.