The Case for a Turing Test 2.0

    

By Jeff Jedras

Actor Benedict Cumberbatch takes high performance computing out of the data center and into the theater with his portrayal of computing industry pioneer Alan Turing in the Oscar-nominated film, The Imitation Game.

The movie tells the story of how Turing led a team during World War Two that helped crack Germany’s naval Enigma code, allowing the Allies to gain a crucial advantage against the U-Boats that threatened to choke off the resupply of Britain. It was a technology victory that, the film argues, shortened the war and saved millions of lives.turing

Perhaps a sequel will focus on one of Turing’s other claims to fame: the Turing Test. Proposed by Turing in 1950, it seeks to distinguish between a machine-based artificial intelligence and a human. In a nutshell, it asks, "Are you a human, or a machine trying to fool me?"

The Turing Test anticipated the day when technology became so advanced that distinguishing between the two could be impossible – the day when machines could actually think.

The advances in the 65 years since Turing wrote his paper have been remarkable, and, as a recent Killscreen article notes, several recent attempts have come close to fooling Turing’s test by sidestepping questions designed to expose human/machine differences, such as contextual questions. For example, Eliza, a chatbot, poses as a therapist and therefore deflects many questions with a follow-up question – just as a therapist would.

That in itself raises a question: do we need a new Turing Test to account for today's technology? The crew at Radio Free HPC tackle that question in their latest podcast (click here to listen). As they point out, technology is already pretty smart. Apple’s Siri, while limited, is a major step forward in human-like simulation. A big focus for IBM's Watson analytics software has been around one of Turing’s red flags: context. Watson's success on the game show Jeopardy is due to its ability to infer the right answers based on context by scanning a range of databases and making a statistical evaluation of which is likeliest to be accurate for the particular question at hand.

One idea the podcast proposes that will likely get more attention as interfaces that mimic humans become more prevalent: a disclosure informing users that they're dealing with an artificial intelligence. Even a new Turing Test may not be able to keep up with new technologies in the years to come.

While advances in processing speed and memory are allowing computers to inch closer to replicating the human brain, there is still one area that they have yet to solve and that may form the basis of a Turing Test 2.0: emotion. Machines may be able to master context. But can they handle the illogic of human emotions? The day they can is the day we should really be scared.

 

Jeff Jedras has written for ComputerWorld Canada, ITBusiness.ca and Computer Dealer News. He lives in Toronto.

--------------------------------

Remember to subscribe to our blog to stay on top of the latest tips, innovations, and best practices for cluster management. 

--------------------------------

High Performance Computing eBook