AI is not at all like Mobile/Cloud/SaaS
AI is a hard problem and will take much longer to solve in any scope. The sudden uptick in interest may revert back to normal, but the cycle of work will be longer, much more diverse, and interesting than Mobile/Cloud/SaaS.
By Alexy Khrabrov, Cicero.AI
There’s an undeniable tsunami of hype in the AI space, and many compare it to the preceding waves of Cloud, SaaS, and Mobile “revolutions”. (Remember “mobile”?). At several AI conferences and VC panels, the question was asked, is AI next, and is it made of the same cloth? Will the AI technology cycle follow the hype cycle, and will it resemble Cloud/SaaS?
On the panels, most VCs said yes. In my informal Twitter poll 75% also said yes. Others said that AI will be like water. The joke goes like: an old fish passes by two younger ones and says, “The water’s good today”. The younger ones look at each other and ask, “what’s water?” Electricity and water are ubiquitous and are assumed to just be there, all around you. So will be AI.
My [definition of AI] is “computer-augmented decision-making in context”. I don’t think this comparison fully holds, however. I agree with the idea that Cloud, SaaS and Mobile are the key enablers of AI, since the data will be amassed together, next to the vast computer required to understand it, and sensing will occur in context. Hence Andrew Ng’s now famous rocket analogy will be satisfied (Deep Learning == vast amounts of data + vast amounts of compute).
However, intelligence is much more diverse that cloud or mobile. Let us ask a simple question: when can you say that “mobile” or “cloud” has succeeded? The answer is quite obvious in each case — when your app works on a smartphone, or when your app runs in a remote data center. But how can we say that AI “succeeds”? When a system is acting like a human? When it can make good decisions? When it can outperform a human on some business metrics?
We live our whole lives evaluating other people’s intelligence. We try to improve ours, too. This is in sharp contrast with the narrow technical challenges of moving something “into the cloud” or making it small or asynchronous enough to be “mobile”. I think that point by itself is sufficient to draw a sharp contrast between AI and other kinds of revolutions that scrolled on the screens lately.
AI is a hard problem and will take much longer to solve in any scope.
The sudden uptick in interest may revert back to normal, but the cycle of work will be longer, much more diverse, and interesting than Mobile/Cloud/SaaS.
Bio: Alexy Khrabrov is the Chief Scientist at Cicero.AI, a community-based research institute, organizing AI By the Bay, a technical conference to define and understand AI, [self.driving.cars], and [ai.vision], to be held in San Francisco March 6-8, 2017.
- Machine Learning: Separating Hype From Reality
- Deep Learning is not Enough
- Avoiding Another AI Winter