June is perhaps a little early to start declaring winners for the whole of 2019. Nevertheless, I’m willing to go out on a limb for a recent article in the Wall Street Journal. The article “When Quantum Computing Meets AI: Smarter Digital Assistants and More” may not end up being the silliest technology article of the year – but I feel confident it will be a strong contender. (As first published in the newspaper, the title was a little different: “Digital Assistants to Get a Boost at a Quantum Level).
In the article, some people gush about the potential for quantum computing to revolutionize artificial intelligence within five years. And what, we might wonder, is the basis for this claim? Is there, for example, an interesting example of quantum computing actually improving the quality of a digital assistant? No. Instead, we learn that quantum computing is very different. It could be very fast because it’s so different. It could be very effective because it’s so different. Well, yes. But lots of things are possible, if we simply mean they’re “not impossible.”
When you look at the article carefully and skeptically, there’s almost nothing to it. One quote – from someone who was just named CEO of a quantum computing startup – identifies frustrations with current digital assistants, and says that quantum computers “could alleviate those problems.” Sure, they could… but they might not.
Likewise, someone from Microsoft says that the potential impact of quantum computing is “a real selling point” – which can be completely truthful with respect to sales technique, but tells you literally nothing about whether the phenomenon is real.
Then there’s someone from IBM who says that “quantum computers can make AI even more powerful.” Sure, they can… unless they don’t. It’s frankly too early to make that call.
The VW experiment
Is there anything real underpinning all the hype and hypotheticals? Yes, sort of. The most concrete item is not related to digital assistants: there’s a mention, toward the end, of some research being done by Volkswagen on intelligent traffic management. The article claims that “the system could be embedded in its vehicles or could be used by public-transportation or ride-sharing companies to predict revenues based on more accurate passenger demand.” I suspect that the latter part of the sentence is garbled – Volkswagen’s more detailed description seems to focus more on better placement of taxis. But it’s worth understanding a little more about what Volkswagen’s research did or didn’t demonstrate.
First, VW took a large data set of mobile phone location data and taxi data. Then they demonstrated that they could use a quantum computer to optimize taxi placement for taxi demand. That is interesting. But there’s nothing in any of the publicly available information about this experiment that claims any improvement over what they could have done conventionally. So unless you are keen on quantum computing for its own sake, it’s not clear that much has really been accomplished.
My, that’s a big box
What about the idea that Volkswagen’s system could be “embedded in vehicles?” It’s worth looking at the public presentation – particularly the first part, where an executive from D-Wave is explaining the quantum computer. The computer itself is quite tiny, but it’s shielded from interference by layer upon layer upon layer. It winds up being a cube that’s roughly 3 meters on a side, with an exotic cooling apparatus to get the quantum computer cold enough. That might be something that could be embedded in a dump truck or an 18-wheeler – albeit at the expense of quite a bit of carrying capacity – but it doesn’t seem like a candidate for embedding in any other vehicle in the foreseeable future.
It’s easy to make predictions, it’s hard to make products
It’s fair to say that quantum computing has considerable interesting potential. But let’s be realistic. We can get a better perspective by comparing quantum computing to blockchain. We’ve known how to run blockchains reliably for more than ten years, but it’s still a challenge to find the right applications for them. In contrast, quantum computing doesn’t even have stable hardware or software to solve some of the problems for which it’s being hyped in this article.
If there were a working version of a quantum-computing digital assistant in a research lab, we would still expect it to take additional years for that to become a product that mattered commercially. To claim a 5-year timeframe for something that hasn’t even been invented yet is the height of technological silliness.