2015 : WHAT DO YOU THINK ABOUT MACHINES THAT THINK?

sheizaf_rafaeli's picture
Professor, Director, The Center for Internet Research, University of Haifa, Israel
Thinking Machines Are About Communication

Thinking machines are not here yet. But they will let us know if and when they surface. And that's the point. To me, thinking machines are about communication.

By thinking, machines might be saved from the tragic role into which they have been cast in human culture. For centuries, thinking machines were both a looming threat and a receding target. At once, the thinking machine is perennially just beyond grasp, continuously sought after, and repeatedly waved threateningly in dystopic caveats. For decades, the field of artificial intelligence suffered the syndrome of moving goalposts. As soon as an intelligence development target was reached, it was redefined, and consequently no longer recognized as "intelligent". This process took place with calculation, playing trivia as well as with more serious games like chess. It was the course followed by voice and picture recognition, natural language understanding and translation. Whereas the development horizon keeps expanding, we become continuously harder to impress. So the goal of "thinking", like the older one of "intelligence", can use some thought. Forethought.

We should not limit discussion merely to thinking. We should think about discussion too. Information is more than just data, by being less voluminous and more relevant. Knowledge goes beyond mere information by being applicable, not just abundant. Wisdom is knowing how not to get into binds for which smarts only indicate the escape  routes. And thinking? Thinking needs data, information and knowledge, but also requires communication and interaction. Thinking is about asking questions, not just answering them.

Communication and interaction are the new location for the goalposts. Thinking about thinking transcends smarts and wisdom. Thinking implies consciousness and sentience. And here data, information, even knowledge, calculation, memory and perception are not enough. For a machine to think it will need to be curious, creative and communicative. I have no doubt this will happen. Soon. But the cycle will be completed only once machines will be able to converse: phrase, pose and rephrase questions that we now only marvel at their ability to answer.

Machines that think could be a great idea. Just like machines that move, cook, reproduce, protect, they can make our lives easier, and perhaps even better. When they do, they will be most welcome. I suspect that when this happens, the event will be less dramatic or traumatic than feared by some. A thinking machine will only really happen when it is able to inform us,  as well as perceive, contain and process reactions. A true thinking machine will even console the trauma and provide relief for the drama.

Thinking machines will be worth thinking about, ergo will really think, when they truly interact. In other words, they will only really think when they say so, convincingly, at their own initiative, and hopefully after they have discussed it among "themselves". Machines will think, in the full sense of the word, once they form communities, and join in ours. If, and when, machines care enough to do so, and form a bond that gets others excited enough to talk it over with them, they will have passed the "thinking" test. 

Note that this is a higher bar than the one set by Turing. Like thinking, interaction is something not all people do, and most do not do well. If and when machines truly interact, in a rich, rewarding, and resonating manner that is possible but rare even among humans, we will have something to truly fret, worry about, and in my view, mostly celebrate.

Machines that calculate, remember, even create and conjecture amazingly well, are yesterday's news. Machines will think when they communicate. Machines that think will converse with each other as well as with other sentient beings. They will autonomously create messages and thread them into ongoing relations, they will then successfully and independently react to outside stimuli. Much like intelligent pets, who many would swear are capable of both thinking and maintaining relationships, intelligent synthetic devices will "think", when they have the ability to convince enough of us to contemplate, believe and accept the fact that they are indeed thinking.  When this happens, it will probably be less traumatic than some expect.

Machines that talk, remember, amuse or fly were all feared not too long ago, and are now commonplace, no longer considered magic or unique. The making and proof of thinking machines, as well as the consolation for machines encroaching on the most human of domains, will be in a deconstruction of the remaining frontier: that of communication. Synthesizing interaction may prove to be the last frontier. And when machines do so well, they will do the advocacy for themselves.