Thursday, May 25, 2017

Advances in quantum computing: presentation by Dr. Brian La Cour

Dr. Brian La Cour from University of Texas at Austin gave a presentation on the latest state of quantum computing to Austin's Quantum Computing meetup in March of 2017. Here are some prominent points from it.

Several big companies are getting into quantum computing now: Google, Microsoft, IBM. There are significant differences between their approaches.

Google plans to solve a problem with 49 qubits, a problem that would demonstrate quantum supremacy (getting a clear speedup with a quantum algorithm over a classical algorithm), but is completely useless in real life. The problem with that is that when you progress to the quantum supremacy frontier, you can no longer check the answer with a regular device. Or it would take a very long time. So Google's argument will probably be based on asymptotic trend, with how they are doing with more qubits. But overall this problem of how to check the results will be more difficult in the future.

Brian La Cour talks at the Austin Quantum Computing meetup
Brian La Cour talks about Google's race to quantum supremacy at the Austin Quantum Computing meetup

We are a long way from solving Shor's algorithm and breaking internet's encryption. But quantum simulation is a near term application. It goes back to Richard Feynman's discussions of quantum computing. We can simulate things on a digital computer, but it's not very efficient. When you want to add another spin, another atom, you have to double the memory. What better thing to simulate a quantum system than a quantum system?

There is a difference between gate-based quantum computing (like what IBM and partially Google does) and quantum annealing, like what D-Wave does, and partially Google.

Gate-based device (that operates on gates, similar to classical gates) is a universal quantum computer. D-Wave's computer is specialized, it is only useful for certain optimization problems. And so far those problems have been pretty contrived, not necessarily corresponding to anything in real life. Even so there is no definitive evidence that the D-Wave's computer is advantageous for solving specific practical problems, as compared to classical solvers. There is a professor somewhere who, every time when D-Wave claimed that their quantum computer was solving some problems more efficiently, took it as a challenge to find a classical algorithm that would beat it. And so far he has been successful. But lately this has become less clear, because he has been, in Brian's words "exploiting what he knows about the problem". (Mathematicians and computer scientists can make it sound like it's a bad thing. But perhaps he means that while the professor is exploiting special knowledge about a problem, the quantum annealing computer can't make use of that knowledge, thus he is not comparing apples to apples? -- E.)

IARPA -- funding agency for intelligence community, analogy of DARPA -- is focused on developing next-generation quantum annealing, a universal quantum annealer. Their goal is, can you take benchmark projects and scale them to the thousands of qubits that D-Wave has?

Microsoft is looking at high-level languages for quantum computing. They are designing high-level languages that optimize what low-level languages do. They also do their own research into topological quantum computing, which is a very different approach than the qubit-based QC, but Brian thinks that's technologically so far away it's probably never going to happen.

This brings us to another Brian La Cour point, which is that now is a good time even for ordinary software developers to get involved in quantum computing, and you don't have to be a researcher to do it. There are people who are building interfaces in conventional programming languages to QASM, IBM's Quantum assembly language. This is where you as an individual can make a contribution: figure out how to do things in QASM and implement an interface to it in your favorite language. Also, individuals can play around with the IBM's Quantum Experience, a web interface to the IBM's quantum computer, and familiarity with it could put you in a position to get a job at some company that does quantum computing (not that there are many of those currently -- E.).

According to Brian, quantum gamification is also a trend. However, he used the word "gamification" not the way it is typically used (to incentivize certain user behaviors by making them seem like a game). He meant it more literally in the sense of games that teach you something about quantum mechanics. In some of those games people perform actions that help quantum researchers. Here are some examples:

  • www.scienceathome.org
  • qCraft -- a "mod" to the popular Minecraft game. It uses "quantum blocks" to teach superposition, observation, and entanglement.
  • There are even games that have been programmed on IBM Quantum Experience, such as Quantum Battleship.
  • Decodoku was developed to help researchers with quantum error correction. You are protecting researchers from errors.
  • cat-paper-scissors game (I could not find it by googling -- E.)
  • Quantum Cats from University of Waterloo, Canada. It's like angry birds, except cats can be in superpositions.

Brian La Cour noted that not only quantum computing research is strong in Canada, Canadian QC scientists also do a lot of educational outreach. US scientists don't do nearly as much, but they should.

As always, the least predictable part of any presentation is the audience's questions, and Brian got a few of those.

Audience member. Have you heard of neuromorphic computing?

Brian replied that he has heard about it, but that there wasn't any connection there to quantum computing. It was just another unconventional way to compute. Apropos of unconventional computing, quantum computing in a way is a throwback to analog computing: encoding information in continuous variables, except what comes out is still digital.

Audience member. Can quantum computing be used to mine bitcoins?

Aside from currently existing quantum computers being nowhere near powerful enough to mine bitcoins, Brian also noted that the value of bitcoin is based on the fact that bitcoins are computationally difficult to find. So if you find an algorithm to mine them fast and reliably, it will devalue them.

Audience member. How big a leap is it to go from classical programming to quantum programming? Is it a totally different beast?

Brian. It is a totally different beast. If you try to do things like conditionals and loops, you are doing it wrong. Instead of conditionals you have control gates, where value of one qubit controls what happens to another qubit. Which is sort of like conditional, but linear. A qubit in a superposition of 0/1 controls another qubit which is also in superposition.

The way you think about quantum computing is taking your entire data space, or state space, and think about it all at once.

You initialize all to 0 and apply Hadamard gates all at once. It puts them in superposition of all possible states, and then you do operation on them. You are looking at the whole haystack and apply operations to the whole haystack until you find a needle. You don't examine each value and look whether it's a needle.

Wednesday, February 15, 2017

Stumbling into BodyHackingCon on the last day

You walk into the BodyHackingCon on an early Sunday afternoon, and you are not sure if it's really still going on. You expect it to have a bigger, or at least flashier presence: shouldn't there be people with highly visible body modifications milling about? Instead, you see people in business casual whose name tags say "superintendent", and the hallways are plastered with signs for Texas School Boards convention. But you persist and walk around a corner, and then down a city-block-long corridor around another corner (that's Austin Convention Center for you), and you are finally rewarded by a hand-scrawled sign pointing towards an open door of a huge, warehouse-style expo room. But this is the last day of a 3-day convention, so naturally most of the vendors are gone.

Neosensory vest that is supposed to let you perceive words as vibrations on your skin, seen at BodyHackingCon 2017 in Austin
Neosensory vest that is supposed to let you perceive words as vibrations on your skin

There is still a thing or two happening; at one of the booths a visitor is trying to pull together the edges of a peculiar-looking vest around his torso; it clearly is not going to happen, since the vest is 3-4 sizes too small. "I'm sorry. We are planning to have larger sizes in the future," says a vendor at the booth, even though the guy is merely average size. But apparently the vest does not need to close to work. It is studded with small metal circles that make up some kind of haptic language interface. That's only my guess based on what I could glean from the snippets of conversation. Because who needs to ask how it works when you can speculate?

"Whip. Angle," says the booth guy. "Whip. Angle." Then he turns a phone screen to the guy who's trying out the vest. There are two circles on it, and he asks the guy to pick one to tap on. Apparently the booth guy made the dots convey some kind of haptic stimulation (e.g. buzzing?) -- and asked the wearer to recognize the word encoded in it. He praises the wearer for answering correctly. "So you see, it's not just the length of the word," he says. I guess he was saying that the vest made it possible, with some minimal training, to distinguish the actual word pattern, not just a longer word from a shorter word?

A jacket with a ribcage headpiece (?) from BodyHackingCon
A jacket with a ribcage headpiece? This would make a splash at a science fiction convention.

Then you look around some more, and even with most vendors gone and large patches of the expo hall square footage reverting to its post-convention beige bleakness, you still see something unusual. At another exhibitor's booth, flanking it on both sides, two women are lying on the tables, looking for all the world like wax statues. Their eyes are covered with something that could be a sleep mask or a VR headset. You glance at the vendor's name -- bio- or healing-something -- and think it's more likely to be a mask. By the way, the name matches a definite pattern: half of the exhibitors' names here have "bio", or "quantum", or something vaguely medical in a New Agey way. Which is fitting, given that half of them sell nothing more than nutrition drinks and supplements.

You stumble upon exhibits of clothes that wouldn't be out of place a goth or punk store, except they have patches with wires sticking out, like something that's placed on you right before a surgery. Some also light up. Many would make a stunning costume at a science fiction convention, if you could spawn off a third or fourth alter ego to explore your mild interest in costuming. Overall, this is the bodyhacking you could get behind -- the kind that stays entirely outside the body.

Most of those clothes are art projects. One dress claims to simulate dark matter: "Dark Matter inflates and deflates against your body to simulate the universe expanding against you, and the buzzing sculptural universal necklace, "Dark Energy", buzzes against your skin to simulate movement through the universe in time in accordance with events happening in VR". But you have read enough science fiction and imagined the vast cosmic space enough times that you know if you put on that dress (not that it's an option) the experience would fall very short of feeling at the center of the expanding universe.

A cape that goes over some sensor with wires that's placed on your chest, seen at BodyHackingCon 2017 in Austin
A cape that goes over some sensor with wires that's placed on your chest, resembling uncannily of surgical preparations.

Some of the clothes have VR content associated with it accessible through your phone; and perhaps you could spend some interesting minutes with it, but just downloading the app would take some time, and the WiFi connection in this building is iffy, and the event is winding down and you are sure vendors are anxious to pack up and leave.

Finally on the way out you get a glimpse of a more radical kind of bodyhacking: a guy you pass in the hallway has small, but prominent devil's horns under the skin of his bald forehead.