In the applications we’ve discussed so far, the multicomputational paradigm enters basically in a descriptive way, providing raw material from which models can be made. In distributed computing, the paradigm also plays a very powerful prescriptive role, suggesting new ways to do computations, and new kinds of computations to do. One can think of traditional sequential computation as being based on a simple chain of “evaluation events”, with each event being essentially the evaluation of a function that transforms its input to produce output. The input and output that the functions deal with can involve many “parallel” elements (as, for example, in a cellular automaton) but there’s always just “one output” produced by a given function. The multicomputational paradigm, however, suggests computations that involve not just chains of evaluation events, but more complicated graphs of them. And one way this can arise is through having “functions” (or “evaluation events”) that produce not just one but several “results” that can then independently be “consumed” by future evaluation events. A feature of traditional sequential computations is that they’re immediately suitable for execution on a single processor that successively performs a chain of evaluations. But the multicomputational paradigm involves in a sense lots of “separate” evaluation events, that can potentially occur on a whole distributed collection of processors. There are definite causal relations that have to exist between evaluation events, but there need be no single, total ordering. Some events require as input the output from other events, and so have a definite relative ordering, making them—in physics terminology—“timelike separated”. Some events can be executed in parallel, essentially independently or “asynchronously” of each other—and so can be considered “spacelike separated”. And some events can be executed in different orders, but doing so can lead to different results—making these events (in our physics terminology) “branchlike separated”. In practical distributed computing, there are usually great efforts made to avoid branchlike-separated events (or “race conditions”, as they’re commonly called). And if one can do this then one has a computation that—despite its distributed character—can still be interpreted in a fundamentally sequential way in time, with a succession of “definite results” being generated. And, yes, this is certainly a natural thing for us humans to try to do, because it’s what allows us to map the computation into our typical sequentialized “single thread of experience” that seems to be a fundamental feature of our basic notion of consciousness. But what the multicomputational paradigm suggests is that this isn’t the only way to set up distributed computing. Instead, we just want to think about the progress of a computation in terms of the “bulk perception” of some observer. The observer may be able to pick many different reference frames, but each will represent some computation. Sometimes this computation will correspond to a distributed version of something we were already familiar with. But sometimes it’ll effectively be a new kind of computation. It’s common to talk about nondeterministic computation in which many paths can be followed—but ultimately one picks out one particular path (say, one that successfully satisfies some condition one’s looking for). The multicomputational paradigm is about the rather different idea of actually treating the “answer” as corresponding to a whole bundle of paths that are combined or conflated through a choice of reference frame. And, yes, this type of thing is rather alien to our traditional “single-thread-of-time” experience of computing. But the point is that particularly through its use and interpretation in physics and so many other areas the multicomputational paradigm gives us a general way to think about—and harness—such things. And potentially gives us a very different—and powerful—new approach to distributed computing, perhaps complete with very general physics-like “bulk” laws. OK, so what about other areas? There are quite a few more that I’ve thought at least a little about. Among them are ones like history, psychology and the general development of knowledge. And, yes, it might seem quite surprising that there could be anything scientific or systematic to say about such areas. But the remarkable thing is that by having a new paradigm for theoretical science—the multicomputational paradigm—it becomes conceivable to start bringing science to areas it’s never been able to touch before. But even in what I’ve said above, I’ve only just begun to sketch how the multicomputational paradigm might apply in different areas. In each case there’s years of work to do in developing and refining things. But I think there’s amazing promise in expanding the domain of theoretical science, and potentially bringing physics-like laws to fields that have long sought such things, but never been able to find them.