Tune in to Timothy Tuttle, the founder and CEO of Expect Labs, talk about a long-standing debate happening in the AI community.
Pair with this video about new advances in the field of deep learning.
TRANSCRIPT:
So, there’s a long standing debate in the AI community about what constitutes true intelligence, true machine intelligence. This debate has been going on as long as AI has existed, going back 50 years it goes back to the days of John McCarthy and Marvin Minsky and certainly was an active debate when I was studying AI at MIT. The essence of the debate really breaks down between the engineers on one side and the scientists on the other side. So, the engineers argue that there are two types of intelligence that can be replicated by performing simple tasks, like pattern matching, at a large enough scale so that when you have the ability to recognize enough patterns, you have the ability to create something that’s seemingly intelligent. And it would argue that the ability for a child to recognize the difference between an apple and an orange is the same as a computer system’s ability to know the difference between an apple and an orange by looking at thousands of examples of apples and oranges and training itself that way.
Now, on the other side of the debate are the scientists which argue that simple pattern matching, while it may demonstrate certain levels of intelligence, does not reflect how the brain actually works. The argument around this is something like you can build a say a computer system to answer a very simple question like what is the capital of France? A computer system would be able to say by looking at databases of questions it’ll probably be able to find an example of that question and tell you that the capital of France is Paris. But if you consider a more complex question like, “can an elephant balance a checkbook?” In order to answer that question you first need to know that elephants don’t know how to do math, and they probably can’t pick up a pen, and the probably don’t own a check book, and putting all those things together you can make a determination that an elephant can’t balance a checkbook. And so, because computers can’t answer questions like that, as well as very simple questions. A lot of scientists argue that true human intelligence is not about simple pattern matching.
I think the other reason that this debate is so popular is because, the field of artificial intelligence has hit a few speed bumps along the way. In fact, the applications that happen to be getting all the attention that would be viewed as intelligent, would be applications like Siri or the IBM Watson supercomputer that can play Jeopardy. These applications are making the headlines, so they get all the attention. Where as, a lot of the true research that’s being done under the hood in the laboratories, doesn’t get a lot of attention. And at the same, because many of these public applications don’t really meet the expectations that people have around what human intelligence should be. In many cases, they can give AI a bad name. So, I think that’s the crux of this debate that’s happening.
