I linked last week to a story about the recently published AI Index
report, but only got around to actually diving into the report today. For the purposes of quantifying progress in artificial intelligence, it’s a good start and something I would encourage anybody interested in AI to take a look at. There’s something inherently useful about trying to visualize progress in a field as fast-moving, and often nebulous, as AI.
Most, of the pieces I’ve seen covering it (including this one from the New York Times
, this one from MIT Tech Review
and this one from Bloomberg
) have headlines along the lines of “AI isn’t yet everything we’ve been promised it will be.” That’s a fair takeaway from the report, but also an oversimplification. While a significant portion of media coverage about AI is, to put it generously, hyperbolic, it’s hard to overstate the capabilities of on very specific tasks (e.g., image and speech recognition) and how they have affected areas such as consumer electronics, enterprise IT and health care.
Even if claims about how AI will affect the economy or remake our day-to-day lives (think smart gadgets and driverless cars) in 5 years don’t come to fruition, there’s enough of a there there to justify all the excitement. The advent of deep learning and its ecosystem of techniques has likely forever changed the way everyone from software developers to CEOs think about what a product can be or which business processes can be automated. I actually don’t expect to see any one profound AI thing (event, discovery, product or whatever) that will fundamentally change the world, but rather a bunch of seemingly mundane improvements that collectively save us time, money, stress, illness, etc. – possibly without many end-users even realizing it.
And this is why I think the AI Index’s most useful aspect is how open it is about its limitations and how open its authors appear to be about receiving feedback on how to improve it. As they and the report’s “expert forum” note, it’s very difficult to quantify concepts such as general intelligence or the impact of AI on any specific vertical market. As important as those types of data are, even more important – and presumably even more difficult to quantify – will be the effects of AI on society, in areas ranging from interpersonal communication to criminal justice.
The AI Index is one of those occasional reminders that even in the era of “big data,” there’s still a lot of stuff we just can’t condense into a single graph or number. Truly capturing the state of certain complex issues takes thinking deeply about the problem in many different ways, rather than simply taking whatever data is easily available and building an argument around that.
China, the United States and the ongoing AI arms race
I’ve written about this a few times, including in these two posts:
and frankly don’t have a lot more to say on the subject at the moment.
But because it’s an important topic and one that just continues to generate news stories and expert commentary, here are a few good pieces I came across this morning:
And, also, in completely unrelated news, this is a fair take from Preston Gralla in Computerworld about how Microsoft’s failure in mobile could make it an also-ran in AI
. I happen to think it’s not entirely true – because, without any data to back it up, I have a gut feeling that consumer AI will actually see something of backlash or bubble burst, and that Microsoft stands to make a mint integrating AI into all things Azure, Office, etc. – but it’s a solid argument to the extent that mobile matters.