A Blog by Jonathan Low

 

Mar 2, 2017

Why IBM's Watson Team Does Not See Value In Replicating the Human Mind

For cognitive systems to be successfully adopted and scalable requires an intelligence that augments rather than mimics the human mind. JL

Frederic Lardinois reports in Tech Crunch:

Cognitive computing is about amplifying human cognition — seeing perspectives you haven’t seen before, being called out when your biases may blind you. The challenge is understanding people, not just what they vocalize. "We already have 8 billion human minds. The kind of intelligence we need to focus on in a cognitive system is a different kind from the one humans possess.” For people to be comfortable takes a system that is highly personalized and can recognize what mood you are in
Everybody and their mother is out to create their own specialized voice-activated devices,” IBM fellow and CTO for its Watson project Rob High told me during an interview at MWC. IBM, of course, doesn’t offer a direct competitor to Siri, Google Assistant or Alexa, but the company hopes that developers will choose Watson, in all its various guises, to power their AI apps, smart chatbots and similar services.
High was not very positive about the current generation of chatbots and virtual agents but he believes that IBM may have the technology to push this far enough so that its users can get more utility out of them. “Classic mobile applications, web applications, IVRs — all those channels are prone to getting advantage to having a virtual agent that has more depth to conduct a conversation than you classically see with Siri or early version Alexa chatbots,” he said. “I don’t even know why they call them chatbots. It’s really command-and-response or single-utterance interactions.”

He noted that frame-based dialoguing (think ordering a pizza with lots of options through a back and forth with the computer) is something the company will launch soon, as well as more advanced cognitive technologies that will allow for the deeper level of reasoning that’s necessary to sustain a conversation.
That’s not easy, though, the biggest challenge here is “understanding people, not just what they vocalize,” High said. “Most of the history of AI has been centered on simple translation and recognition kind of techniques. That’s why deep learning has become so popular. It’s really good at doing recognition. But that doesn’t give us a deep understanding about a person’s motivation. Our motivations are often not expressed upfront, yet it is key to understanding what people want to get done.”
For High, this isn’t about replicating human intelligence, though. “Cognitive computing is about amplifying human cognition — whether that is in the context of a conversation or whether that is more of in a discovery scenario where the implicit goal is to come up with new ideas, seeing perspectives you haven’t seen before, seeing through your biases, being called out when your biases may dead-end you or blind you to things.
I don’t believe there is a lot of value in replicating the human mind in a computer.
— Rob High
“We, as human beings, should be rightfully proud about the intelligence we possess,” High said but also noted that our intelligence is constrained by how much we can read and retain, for example. “So if we can take that and augment it using a cognitive computer that doesn’t replicate everything we do in our own mental process but analyzes, augments and amplifies it, that to me is the ideal — that’s the goal for cognitive computing for augmented intelligence.”He believes that this will result in our ability as humans to fully realize our energy and smarts. “I don’t believe there is a lot of value in replicating the human mind in a computer. We already have 8 billion human minds. The kind of intelligence we need to focus on in a cognitive system is a different kind of intelligence from the one that humans possess.” He added that we don’t really know what that kind of intelligence will look like, though. “Our lifetimes are not the right measurement for the history these systems will create for themselves.”
The best way to bring this intelligence to users still remains to be discovered, though. High stressed that it takes a sense of “presence” — maybe through a robot or built into anything from a TV to your phones — for people to be comfortable with a system like this, though. That takes a system that is highly personalized, too. A system that can recognize what mood you are in to play the right music or that automatically turns down the music when you start a conversation. “That’s where we are going, but that’s beyond where we are today — and not just translations,” High added.
High believes that Watson’s ability to do higher-level reasoning (at least to some degree), puts it beyond its competitors’ capabilities today. He acknowledged that there will always be some commoditization — “and that’s ok because we’ll just continue to advance.”

0 comments:

Post a Comment