Think about the word goal. Now think about a goal that you have. Can a robot endowed with artificial intelligence have one, too?
MIT's Josh Tenenbaum explored the idea in a panel discussion at the AI and the Future of Work conference on the school's Cambridge, Mass., campus earlier this month. Describing himself as "half AI researcher and half cognitive scientist," Tenenbaum studies the mechanics of how humans think and then uses that understanding to build smarter, more humanlike machine intelligence.
A goal is one of the things that separate humans and machines today, he said. Humans have goals, ambitions, things they strive for -- going to law school, mastering the guitar, becoming a computer engineer. A machine such as DeepMind can be the best Go player in the world, but it can't decide it wants to be good at Go and "that that's actually what it wants to spend its time doing the way a Go champion does," Tenenbaum said.
But goal is a "suitcase word," said fellow panelist and MIT professor Patrick Winston, an AI and computer science researcher. It's a term from AI pioneer Marvin Minsky for a concept that seems direct and clear but is really endlessly complex -- packed with many ideas and processes and problems, like a suitcase is packed with clothes and travel gear. For example, people use the word consciousness, Minsky wrote in his book The Emotion Machine, for more than 20 mental processes, including learning, recollection, expression and narration.
Winston's point was that until we humans can unpack suitcase words -- goal, consciousness, self-awareness -- and understand more about "what goes on inside our skulls," we're unlikely to build machines that are smarter than we are.
"I think we can flatter ourselves to say that we are beginning to understand how machines can have those kinds of things," he said.
That got me thinking about the universe of catchphrases CIOs have to first make sense of and then take action on -- digital transformation, the cloud, multi-cloud. They look and sound like things you can pick up and turn on, sets of tasks you can just do. But they're stuffed with discrete problems that need solving before they can be more than words. Are they suitcase words? Even if Minsky didn't extend his label to IT buzzwords, they sure do look like them.
Lugging suitcase words in IT
Take the first in my short list above, digital transformation, using digital technologies to wholly reimagine business. CIOs take up digital projects -- like building a mobile app or automating a business process -- but as Gartner analyst Peter Sondergaard said at Symposium/ITxpo in Orlando, Fla., in October, "Digital projects are not a digital business." Becoming one requires not just investments in technology, but in the people who will build and run it. It involves building a strategy for rollout, hand-in-glove alignment with business goals to determine the problems that need solving and then deliver value for customers.
That's a lot to carry for two words.
Cloud computing has spawned a horizon of terms that can be classified as suitcase words. Deciphering the different modes of cloud -- cloud infrastructure, cloud applications, cloud software development platforms -- is a task unto itself. Then there's cloud-first (choosing cloud for new initiatives unless there's a reason not to) multi-cloud (relying on an assortment of cloud services) and even, as I heard at Gartner Catalyst in San Diego in August, "multi-cloud first."
Again, these are words that involve real challenges that organizations need to tackle in ways that make sense for them. When I asked Kinit Salvi, director of enterprise architecture at Sysco Corp., how he's managing the patchwork of cloud services in his computing environment, he went back to a good place: the basics.
"Multi-cloud is good as an end state. We're trying to get into the cloud first," he said.
Talking problems, solving problems
Artificial intelligence, too, has become something of a suitcase word. Organizations get dazzled by technologies like IBM's Watson that can talk and learn and solve big problems -- then they want to go and build something like that, said independent technology consultant Neena Buck at MIT's AI and the Future of Work. But that's the wrong way to go.
"What is the problem that you're trying to solve?" is the question organizations should be asking, Buck said. But that's not happening in the new groups formed to develop AI. "They're not looking at problems to be solved. They're looking at, 'How do we apply this new AI technology?'"
Suitcase words aren't necessarily bad -- whether they describe love, feelings or emotions, as Minsky did, or the sweeping changes to business and commerce brought on by digital technologies. In fact, they give people a way to talk about complicated ideas.
But back to Winston's point, oversimplifying knotty problems could lead to an inaccurate measure of our progress in solving them. For CIOs, who are hired as problem solvers, the miscalculation could cost them big.