After reading a recent ScheerPost column by Michael T. Klare, professor emeritus of peace and world-security studies at Hampshire College and senior visiting fellow at the Arms Control Association, my first thought was, ah, yes, the warmonger maniacs are really pushing as hard as possible for global extinction now.
Professor Klare (https://scheerpost.com/2024/02/21/emergent-ai-behavior-and-human-destiny) describes enthusiasts (maniacs in my vernacular) at the Department of Defense (specifically its DARPA research arm) and their partners at Microsoft and Google as fostering the development of autonomous swarms of robots driven by artificial intelligence (AI) algorithms that communicate as a whole entity, making decisions independent of humans. Such advanced algorithms would “enable autonomous combat systems to communicate with each other and ‘vote’ on preferred modes of attack”. Such software would thus show similarities with biological systems that exhibit swarm behavior, such as the swarming of bees, ants, even wolf packs.
Most disturbing is the description of the propensity of these interconnected AI systems to demonstrate novel, unplanned, therefore unexpected outcomes, described as “emergent behavior.” Reading this passage made me almost physically ill. As a molecular biologist who has steadfastly moved away from the untrammeled reductionism of my discipline toward the theory of CASoS (Complex Adaptive Systems of Systems), I immediately experienced the intellectual underpinning for my gut-level revulsion: an idea cited in CASoS theory to account for positive creative aspects of living systems was now being applied to AI warrior systems in describing how they might arrive at novel and unanticipated modes of destructiveness.
For example, the emergent properties idea is useful in describing how a system that is deterministic at some levels becomes nondeterministic as a whole CASoS, it’s behavior describable only in terms of probabilistic outcomes. Brain neurotransmitter systems like those employing dopamine, serotonin, etc.—mostly familiar to nonscientists because of the drugs that impact them—can be readily described deterministically, their underlying cellular and molecular processes well understood. But the brain as a CASoS displays properties irreducible to the properties of any of its biochemical neuronal systems or any simple numerical combination thereof. Putting aside the vastly complex topic of consciousness, it’s enough to note that brains contrive novel solutions to problems, solutions that they have never before used, and which are therefore unlikely to be somehow stored as memory. Creative insights, new ways of thinking about something in our lives are an undeniable feature of both our humanity and of problem solving by other animal species.
To stretch this a bit further, I always marvel that with only medieval church music, mostly chants, preceding him, J.S. Bach was able to compose the remarkable counterpoint concertos which have moved my soul from the first moment those sounds impacted my eardrums. How did that kind of creativity spontaneously arise without predecessors? Even Mozart, whose music I love and have probably emulated to some extent, had the tonalities of Bach (as well as Vivaldi, Handel, and Haydn) to unavoidably train his neural networks. As a CASoS, Bach’s brain had access to a novel strain of creativity that we might call an emergent property of complexity.
And yes, there are also no deterministic predictions possible about the precise method a wolf pack will use in a specific hunting situation while downing its prey. In addition to each wolf’s brain, the pack as a whole entity is a CASoS, demonstrating creativity as an emergent property of the complexity entailed in the communicative interactions of its individual members. This is likewise exemplary, and unfortunately much closer to the model chosen by DARPA developers than the fantastic counterpoint surges of Bach concertos.
I obviously feel quite strongly about this co-option of the creative properties of complex biological systems—which took billions of years to emerge, constructing the billions of species and the complexity of their interactions in the remarkable biosphere supporting human life—and the use of such complexity for destructive, rather than creative purposes. While it makes sense that, at some level of complexity, we should expect emergent properties from cyber systems, to think about this occurring in the context of weaponry and war is beyond chilling. I also feel strongly about the hubris of individuals who deem themselves important to this perverse initiative in service to what has essentially been/become the most destructive nation ever known on planet earth. Look in the mirror, computer scientist, profess allegiance to your species, to your co-evolved brethren species, to the survival of all of it. Divest of this need to saber-rattle for the dying empire that seems to want to destroy all it can on its way out of existence. Find a different creative path. Peace. Please!