For me the most interesting bit was Scott's "10 second" hypothetical reply to the reporter/general audience on what quantum computing is (found in one of his comments):
"I say something about how a QC is a proposed device that
would solve certain specific problems much faster than we
know how to solve them today, by taking advantage of
quantum mechanics, which generalizes the laws of classical
probability.
Then I talk about how you’d never talk about a -20% chance of rain tomorrow, but quantum mechanics is based on numbers called amplitudes, which can be positive or negative or even complex numbers.
And how, if an event can happen one way with a positive amplitude and another way with a negative amplitude, the two possibilities can “interfere destructively” and cancel each other out, so that the event never happens at all. And how the state of a QC with (say) 1000 bits would have one amplitude for each of 21000 possible settings of the bits—an astronomical amount of information, if one wanted to write it down classically, for example in order to simulate what the QC was doing classically.
But about how, when you measure the QC’s state, you just see a single random output (with its probability determined by its amplitude), not the gargantuan list of possibilities. And about how the goal, in QC, is always to choreograph things so that the possible paths leading to each wrong answer interfere destructively and cancel each other out, (say) some having positive amplitudes and others negative, whereas the paths leading to right answer reinforce.
And how this is a very weird and specialized capability—it’s not nearly as simple as “trying all the answers in parallel” (if you did that, you’d simply observe a random answer), nor is it just a smaller or faster version of ordinary computing (a QC might even be “bigger” or “slower” than an ordinary one; all the hoped-for advantage comes from the QC’s ability to create interference patterns).
Finally I talk about how a QC is known to give huge advantages over any known classical algorithm for a few tasks of practical importance (quantum simulation, breaking almost all the crypto used today…), and it might also give some advantages for broader goals like optimization and machine learning, but that’s an active research topic, and if the advantages exist they’ll probably be more modest and/or specialized."
"I say something about how a QC is a proposed device that would solve certain specific problems much faster than we know how to solve them today, by taking advantage of quantum mechanics, which generalizes the laws of classical probability.
Then I talk about how you’d never talk about a -20% chance of rain tomorrow, but quantum mechanics is based on numbers called amplitudes, which can be positive or negative or even complex numbers.
And how, if an event can happen one way with a positive amplitude and another way with a negative amplitude, the two possibilities can “interfere destructively” and cancel each other out, so that the event never happens at all. And how the state of a QC with (say) 1000 bits would have one amplitude for each of 21000 possible settings of the bits—an astronomical amount of information, if one wanted to write it down classically, for example in order to simulate what the QC was doing classically.
But about how, when you measure the QC’s state, you just see a single random output (with its probability determined by its amplitude), not the gargantuan list of possibilities. And about how the goal, in QC, is always to choreograph things so that the possible paths leading to each wrong answer interfere destructively and cancel each other out, (say) some having positive amplitudes and others negative, whereas the paths leading to right answer reinforce.
And how this is a very weird and specialized capability—it’s not nearly as simple as “trying all the answers in parallel” (if you did that, you’d simply observe a random answer), nor is it just a smaller or faster version of ordinary computing (a QC might even be “bigger” or “slower” than an ordinary one; all the hoped-for advantage comes from the QC’s ability to create interference patterns).
Finally I talk about how a QC is known to give huge advantages over any known classical algorithm for a few tasks of practical importance (quantum simulation, breaking almost all the crypto used today…), and it might also give some advantages for broader goals like optimization and machine learning, but that’s an active research topic, and if the advantages exist they’ll probably be more modest and/or specialized."