The Future of AI
When people ask me, “Will we ever achieve Artificial General Intelligence (AGI)?” I tend to equivocate.
I delve into an abridged explanation of narrow, general and super intelligence, presenting a well-rounded argument for each dominant viewpoint and leaving it to the inquisitor to choose a path of optimism or skepticism.
But equivocal answers are never satisfying. So, I rephrased the question and demanded an answer of myself.
In 10 years, will we look at AI as we do the Segway (a novelty that, once an emblem of the future, is now more of an innovative oddity more akin to a party trick than a globally-embraced technology) or as we do the internet (the ultimate globally-embraced technology upon which we are all inherently reliant)?
Until recently, I’d have chosen the former.
Why?
First and foremost, I equate the internet with a sort of general intelligence — capable of spanning and scaling dozens, if not hundreds, of tasks with equal success.
Today’s AI capabilities are disjointed. We are very much in the narrow, and far from the general.
Granted, we have put ourselves in a difficult position in a strictly semantic sense.
When I think ‘AI’, I don’t first think linear regression, Neural Networks, or even recommendation engines, because those, although certainly novel and impactful, are relatively well established. Meanwhile, areas like autonomous transport, telemedicine, and those wildly unsettling Boston Dynamics robots, are fanciful and fantastical — AI-esque — if only because they have yet to be achieved.
We’ve all heard some version of, “It’s Artificial Intelligence until it starts working, then it’s just technology,” right?
Nevertheless, the aforementioned caveat never seemed enough of a reason to expect AGI, for even if we were to accept all recent technologies as ‘AI’, their interoperability is still lacking, and that is the main differentiator between ANI (Artificial Narrow Intelligence) and AGI.
Upon further reflection, however, I noted an interesting trend:
Most of AI’s existing capabilities are born of necessity (as are most new technologies), but the field of research is so small that only a relative few solutions (usually those with the most funding) are developed.
In other words, it’s an issue of supply and demand. The reason AI seems so disjointed, expensive, and obscure is because AI is disjointed, expensive, and obscure, but only because we have constrained ourselves to certain areas, not because of ultimate technical capabilities.
Think about where we see the most heavy investment in AI today: banking, healthcare, cyber security, autonomous vehicles, IoT, and media.
Yes, one reason there has been more adoption in these areas is because AI lends itself well to these mediums. But, AI lends itself well to these mediums because these mediums have driven the development of AI over the last few years.
It follows, then, that the reason we haven’t seen any truly capable, affordable, explicable AGI is not because it can’t be done, but because there has not been reason (and funding) yet to pursue it.
Alas, there is reason to be optimistic, if only because there is no real cause for pessimism!
So, will AI go the way of the Segway, or the way of the internet? That depends on us. There is no reason to suspect it is technologically infeasible, and if we find reason (and thus funding) for true AGI, I have faith we can get there.
