Monthly Archive for February, 2016

How Soon Will It Happen?

One day, it is sure to happen. One day artificial intelligence [AI] will become so effective that it would be able to perform a large measure of professional work. Of course, professional workers have been steadily losing part of their work to software for many years. Hundreds of applications, including spreadsheets, contract templates, valuation models and computer-assisted design, have chipped away at the tasks of knowledge workers. This effect has been basically hidden as the total demand for professional work of all kinds has soared, driven by the demands of a technologically complex and global economy. In the absence of today’s computer aids, there would have been a severe shortage of professional workers. And this moderating effect continues as the world’s work shifts to knowledge workers. Still, the threat, or the promise, of AI grows in force.

As the ability of AI to learn improves, it will inevitably take over more of a professional’s typical work. And it is certainly the case that machine learning is advancing, and that the pace of its development is increasing. The issue we therefore face is clear: will AI’s absorption of professional work proceed at the sedate pace of the past, allowing society adequate time to adapt; or will machine learning accelerate to produce a seriously disruptive effect? Or perhaps we are looking at a catastrophic effect? The latter effect is far from certain, but it is not implausible.

There are two schools of thought. One holds that AI has now developed most of the fundamental insights it needs to take machine learning to a level that rivals most applications of human thought, or even to exceed them. Indeed, with deep data mining, it is already impossible for a human to process effectively the volumes of data available. As a result, it is argued that only evolutionary advances in AI, together with increases in cheap computing power, are all that is required for a radical transformation of the labour markets. Therefore, the only real question is when, not whether. And the implied answer is sooner, rather than later.

Let us note that autonomous vehicles are moving forward faster than anticipated primarily because of advances in AI. We must ask whether this is a harbinger of much more disruptive effects.

Alternatively, it may be argued that AI is still in need of further breakthrough research, whose outcome is far from assured and whose timing cannot be anticipated. Moreover, most of today’s AI applications are still very routine and are far from high-level analytic thinking. For example, data mining is not much more than brute pattern matching and introductory statistics. And while the ability to drive a car is impressive for a machine, it is completely unimpressive for a human. Any 16-year old can drive a car without crashing it too often. Turning that kid into a lawyer is a rather different matter.

[We omit comment about pure science fiction “singularities”. Speculations are, of course, valid exercises in honing one’s imagination. It is not the basis for prediction, or career planning.]

In light of these uncertainties, ones which continue for some time, each of us must decide why we are better than the machines, or how we could quickly become better if the need arose. Instead of ignoring the possibility, let us examine it.

We should ask if it is possible for a machine to become human, or almost human in its thought processes. Here the answer is clearly and aggressively that we do not know. Any assertion to the contrary is unfounded. The reason has nothing to do with what technology might be invented. Rather, we do not know what it really is to be human. If we are wondering whether a machine can mimic us, we would need to know what we are.

Yes, we are a calculating machine. Yes, we have memory and processing power. Yes, we react to our environment and we understand the electro-chemical basis of emotion. But we are still a mystery to ourselves. Yes, we are hard wired to get angry or sexually aroused. But what makes us look to the sky and imagine infinity? What makes us imagine what we have never seen and never will? What makes us know we are here? What makes us wonder what our destiny is, or if we have one? Where do these ideas come from?

Speaking solely for myself, I am more than an inference engine. What are you?