OK, I *so* don't buy today's comic. To the point of being mildly angry.
There's no denying that there are many, many gaps in human knowledge, but technology does not work like that. Nor does science, really. The comparison to using fire only works on the surface. Let me try and explain why.
I can believe that AI could emerge without us really understanding why and how. The thing is, we still don't really know where to look for consciousness and we have only a fairly rudimentary understanding of how brains work. But, this is mainly due to neural patterns being very complex systems. We can't understand the big picture, because there's just so MUCH of it. We still understand a *lot* of the basic building blocks. We know the chemistry of the brain and many details of it. We understand the physics and chemistry behind neurons firing. We may not have mapped out everything that happens and so we do not *understand* some of what is happening, but we understand *what* is happening.
The thing is, with something like a power source, there's no real room for that. Assuming AI use electricity and not some kind of new, incredible energy beyond the understanding of current physics, we understand the building blocks of matter pretty well. We know what electrons do and how they flow, and there's only so many ways you can induce an electric current. If a battery uses a chemical process, we can trace the chemistry back to where the free electrons come from. If there's a physical process, there needs to be some sort of interaction between electromagnetic fields, and again, these may be complex to a person intimidated by physics, but not to someone who studies these things. The science behing electricity sources is more than a century old.
On top of that, engineering usually *follows* theoretical science, not the other way around. Usually the process is well understood way before we can make it viable. The first atomic bomb operated on a fairly simple principle. Take two particles of a certain kind and smash them together. The problem was to get the particles to *do* that, and that cost a lot of time, money and thinking.
There is no conceivable way anything short of an alien technology operating on unknown physics would utilise something not analyse-able. Complex technology can't be done blindly, most of the difficult advances in technology are still ways to do something in a way that works within the material and energy limitations.
If there's an AI that creates something that current physics can't account for? Fine. But that would still not explain how this thing can be manufactured. If there's an AI that can create something, but can't convey the principle to scientists in English (or whatever)? Fine. The AI may be lying about any of the above? Fine. But you don't go and create a piece of efficient technology accidentally. That's certainly not how things work. Our understanding of physics may be limited, but we're not cavemen, and it's usually technology chasing science, not the other way around. And again, power sources are, in principle, simple things. Theoretical complexity is NOT the limiting factor behind limitations on power source capacity. Actual material and chemistry and weight and size constraints are.
On top of that, if an AI created a device that no science can adequately explain, there's ten different reasons why this special device would not be crammed into consumer electronics. If it truly operates on some principles indistinguishable from magic, they might explode and create a gateway to another dimension, for what we know. Or give people cancers. Or steal souls. Or whatever. If we rule out those possibilities with a high degree of confidence, then we OBVIOUSLY know, at least roughly, what the device does. And that takes us back to actual electromagnetic fields, which again, interact in well-understood ways.
And on top of *that*, if there was a device that only some select AI understood, or perhaps do not understand as well, its existence would have a much, much bigger impact on science and technology and pretty much everything. Starting with the Industrial Revolution, the only limiting factor on what we can globally do has been pretty much the level of available energy. Look at how difficult space exploration is just because we can't get around the very simple problem of fuel. If what is essentially a magic box of energy existed, that'd change a *lot*, in many respects. I know QC has some suspension-of-disbelief-stretching elements, but a special super-power-cell based on, for all intents and purposes, magic is what breaks my suspension of disbelief where Pintsize and May did not.
Oh, and regarding the coffee example - we may not be able to trace every, well, trace element in coffee that contributes to flavour, but that's a matter of practicality. We still know what the main contributions to the taste and scent are (that's pretty much how we can make artificial flavourings of just about anything), and we understand the chemistry of ingredients such as caffeine on the human body pretty well, too. Not knowing every single component in a brewed coffee is nowhere near the same ballpark as "we have a power cell and have no idea how it works". I'd buy a power cell that has some weird side effects that are difficult to map, but not this. And the comic says the AI do not *know* how the technology works, not that there are some fiddly, not well-understood bits. Again, you can't create a device by accident, unless your setting incorporates mad science (which is for all intents and purposes a form of magic).
There's, of course, also the possibility that Clinton just has limited knowledge about the subject or is exaggerating. I see no other explanation that hold much water. That, or I have to accept the poetic license of the world, which I'm prepared to do, but the comic that tries to be exposition-y and creates more issues than it solves kinda leaves a bad taste in my mouth.