In my last post, I claimed
There is still hope on our horizons, particularly when you explore them with an intelligent teamMountains Beyond Molehills: Recovering the Lost Art of the Mathematics of Empathy
Such a statement begs the question: “What is an intelligent team?”
There are all kinds of definitions of “intelligence” depending on who you are talking to and the context of the conversation.
Without getting deep into that, I’d instead like to orient the conversation around two types of intelligence that I believe to be extremely promising for the future:
Swarm Intelligence is a subset of collective intelligence dealing with the behavior of decentralized, self-organized agents that interact with one another over time. Democracies sometimes come to mind, or economies.
Collectively, human beings don’t seem all that too intelligent at times, so I think swarm intelligence can be a fascinating form of intelligence to study as it relates to our own behaviors.
Swarm-intelligent behavior can result from a number of different mechanisms and interactions. For example in the context of ant colony search, ants communicate with one another indirectly via stigmergy. These ants are able to use simple cues in the environment & basic algorithms to cooperate with one another in order to achieve common goals (despite the fact that none of the ants has any specific goal in mind or the means with which to evaluate such a goal even if it did).
In other models of swarm-intelligence, the agents may communicate directly with one another with full knowledge of the system (for example, fish school search) or partial knowledge/random search behavior (for example, artificial bee colony and its variants).
While related to animal intelligence at times, swarm intelligence refers to a kind of “sociological behavior of groups of agents,” in a way. There are, of course, individual forms of animal intelligence that can be extremely fascinating as well.
One very obvious example is man’s best friend: the dogs. In some cases, dogs are more capable of detecting illness than our best medical equipment. One well known example of this is that dogs can often be extremely intelligent at predicting seizures and instructing their owners to lie down.
Not all such forms of animal intelligence are so benevolent. During the Cold War, Russia and the United States deployed trained animals of war against one another. Military dolphin training has made a comeback lately, and the Smithsonian has even gone so far as to write an article that begins with how “The CIA’s Most Highly-Trained Spies Weren’t Even Human.”
Despite their often great intelligence, there are times when animal intelligence backfires. Elephants, for example, have been shown to exhibit PTSD. As poaching and other environmental stresses have disrupted their societies and confined their living spaces, young male elephants have increasingly been engaging in previously unobserved violent & aggressive behaviors.
Thoughts for the future
Interaction networks & stability
In a recent paper by Oliveira, et al, the analytical concept of an Interaction Network is introduced as a means of studying swarm algorithms by abstracting away everything except for the structure of the social interactions – who is passing information to whom, when, where, how, etc.
I believe this is a particularly promising way to study swarm algorithms, as it also enables additional research into the behavior of multi-agent swarming systems.
For example, in the video below where dolphins and birds are collaborating to hunt fish, what does “intelligence” look like? Suppose a fishing boat arrived. At what rate can an intelligent algorithm adapt to changes and the introduction of other intelligent algorithms?
This is a hard problem that requires a lot of simulation and non-linear dynamics, but it is a good problem, and I expect the concept of Interaction Networks to be very helpful as we continue to explore the stability of collective intelligent behavior, for example. I’m curious how the mathematics of symmetry may be applied to the study of such systems (it seems doable to me, and very beneficial as it would enable SQL as a evolutionary systems domain modeling language).
Combinatorial synesthesia, colors, & cognitive non-isolation
Broadly speaking, I am optimistic that synesthesia may be induced via physical activity (tap dancing, say, or singing) and that such “combinations of senses” could be useful for developing languages and tools with which to reason about extremely complex interactions and systems. A great example of that I learned about recently came from Signal Sciences, and their use of color to evoke appropriate responses from human operators.
Alongside wider and more thoughtful use of synesthesia in consumer labels, I think the days of professional specialization are over. AI can learn directed tasks faster than any human being could dream. AI will be coming soon to a neighborhood near you and it will eat your organizational silos for breakfast before you’ve even woken up to eat breakfast. I expect that liberal arts and other forms of cognitive non-isolation will increasingly be an asset in a world increasingly run with AI.
With these changes, I expect the the conventional wisdom of “more data beats better algorithms” to be completely turned on its head.
I am specifically bullish that the Rust language will continue to gain adoption in distributed and embedded systems. Likewise I believe that the market for TinyML and AI-enabled microcontrollers will grow very rapidly.
Likewise, I expect the market for simulation architecture and platforms (such as Hash.ai) to grow considerably, and that they will prove incredibly helpful for studying interaction networks & other complex systems.