Why animals and smart machines can't measure up to human intelligence

If crows can use sophisticated tools and solve problems that require reasoning and IBM's Watson can beat human competitors on Jeopardy!, why do we humans think we're so smart?

That's a question Canadian philosopher and cognitive scientist Paul Thagard explores in his new book. 

In Bots and Beasts: What Makes Machines, Animals and People Smart?, Thagard evaluates the similarities and differences between humans, animals and smart machines.

Thagard spoke to Spark host Nora Young about how things we don't necessarily equate with pure rationality, like social intelligence, emotions, creativity and dreams, also play an important role in human intelligence.

Here is part of their conversation.

Why are humans so intelligent from a biological or evolutionary perspective?

It's partly because we've got large brains. But it's not just size, because you've got some animals like elephants and whales, who have larger brains than us. They have more neurons as well as more space, but our brains work really well, we can do some very complicated things. 

So obviously, one of them is using language, but part of language is being able to think recursively, to have thoughts about thoughts. For example, I know that you know that I know that we're having a conversation. And this kind of recursion seems to be really special to human beings.

There's also quite a fascinating part that comes up in the book, which is just the invention of cooking and the role that that had, can you tell me a bit about that?

That's a great story from a Brazilian neuroscientist. She's the one who figured out how to count the number of neurons in human brains, she invented something she called 'brain soup', where you basically puree brains, and [that] provides a way of counting the number of neurons. 

But she's also come up with some really interesting ideas about how cooking was important, based on some ideas previously come up with by an anthropologist. The basic idea is that the reason we managed to have big brains is that cooking provided enough energy to do it. Because cooking really provides very concentrated forms of calories. 

And so if you don't have cooking, you've got to be running around chasing food or gathering vegetables, and that takes a lot of energy. But once you can cook things, you can get much more efficient forms of energy. 

So one of the things that characterizes human intelligence, as you alluded to, is collaboration, right? We work together to accomplish things. And increasingly, our technologies like writing and the internet allow us to share and cooperate. So what kind of advantage does that give us?

So much of what people do well, we do by working with others. We really depend on what we can do with other people. And there are just huge differences between people and other animals.

You find simple kinds of collaboration among animals, for example, hunting together. But the evidence suggests that they're just doing their own thing. They end up kind of working together because they're each pursuing their own goals, whereas human beings — and it starts with two year olds — are thinking about what's going on in the minds of other people. They're looking to see what other people want, how they can help them and how they can cooperate with them.

So collaboration is crucial for people. And it requires both the recursive ability to think about thoughts, but also requires emotions, because you've got to care about people and you've got to care about them caring about you, in order for collaboration to really work. 

Now you can have different computers working together solving problems, but they don't have the same kind of motivation to do it. Even though they've got that kind of communication, they don't have nearly the level of caring about collaboration that makes so many human social tasks so productive.

As you point out in the book, machine learning has achieved some incredible things (IBM Watson for example). But what are the barriers to machines achieving something that we might really recognize as intelligent? Not Commander Data level stuff, but an understanding of the way the world actually is?

Driverless cars are already a sign that computers can operate in the world. They're not nearly as good at driving as people, but they're really quite impressive. I don't think they're up to a Canadian winter yet, but they're certainly doing fine in places like Phoenix, and they navigate really quite well. 

They've already got some understanding of the world. And it comes not just from people programming them, it partly comes from machine learning. They're able to pick this up because they've got really good sensors that tell them what's going on in the world. And they've got really good learning algorithms that enable them to develop some concepts on their own.

So it's already there in some rudimentary form, but it tends not to be combined with the higher level verbal things. Programs like IBM Watson are really good with words, and there are robots that are good with interacting with the world. Nobody yet has combined the two, where you can have a high-level statistically and verbally sophisticated machine that also works in the world. Once that happens, you're going to get something closer to human-level intelligence.

Written by Samraweet Yohannes. Produced by Nora Young and Samraweet Yohannes.

This Q&A has been condensed and edited for length and clarity. To hear the full conversation with Paul Thagard, click the 'listen' link at the top of the page.