As I was finalizing the syllabus of my new course
Quantum Integer Programming (47-779)
scheduled for Fall 2020, it was fortuitous to participate in the PQI Distinguished Lecture by Hideo Mabuchi (HM):
Coherent Non-Linear Dynamics and the Physics of Computation.
It turns out that HM has a fairly serious hobby:
This triggered a thought of a Picasso (ceramic, not painting) that I had obtained from Christie’s a few years ago, and his quote that seems particularly apt to describe my quantum computing journey:
I begin with an idea and it becomes something else.
Beyond ceramics, of particular interest to me was this question that HM posed in his pre-talk interview (with Jeremy Levy):
Is the “ideal qubit” abstraction based computing model, paralleling the “bit”-based digital gate model, even the right one for quantum computing?
You may recall my previous post What would Steve Jobs do? where I expressed the need to encourage alternatives to Gate-based (and supercooled qubit driven) hardware, and the associated conceptual computing (“circuit”) model.
This was my motivation for creating the 2020 Tayur Prize.
My own brief experience in quantum computing has been to work with an alternative framework, with Ising model as the mathematical abstraction.
Most illuminating (no pun intended!) was this “maximally inverse” insight I obtained from HM’s talk:
Different from classical methods (such as Simulated Annealing) and quantum methods (such as quantum annealing), where the search for a global minimum of a very complex objective function is attempted from above, that is, from solutions that have a higher objective function value, through a series of steps, that either climb over some hills or tunnel through, CIM, in stark contrast, hopes to emerge at the minimum, from below, on an upward bifurcated trek!
Although I frequently delight in things that have no practical point (like purchasing Picasso’s ceramics), with quantum integer programming, I have been looking for “heuristic good solutions” for certain sub-classes of structured optimization problems that arise from industrial applications, rather than just aiming to prove (or disprove the possibility of) “exponential speedup to obtain optimal solution” for stylized models, which, I was happy to hear, is also in line with HM’s perspective on how we may advance into the future with new forms of computing.
Finally, something else I am very excited to continue to think about, encouraged by a remark HM made towards the end of his talk, is the question about hybrid signal formats, that is:
how best to interweave optical (laser) and electronics (FPGA) components,
for the purpose of computation.
Rephrasing using an artistic term that Picasso would have likely approved (hint: Ballets Russes):
A photon-electron pas de deux. 😏