by James Bailey |

There is no heavy lifting in the Week One introduction. We are simply introduced to so-called "supervised learning." Supervised learning builds on a corpus of examples where the correct answer is already known. The Deep Learning algorithm tries to match the approved answers, in the same way that students have done for centuries, as noted by Leonardo da Vinci:

"

In imitable sciences the student can attain equality with the master and can produce similar fruit [such as] mathematics where the pupil takes in as much as the master gives ... like letters where the copy has the same value as the original reproduced indefinitely as is done in the printing of books.

"

It should be obvious that the "imitable sciences," what we now call the exact sciences or the number sciences (think algebra and mechanics) will be the first to be taken over wholesale by Deep Learning. They are the low hanging fruit.

Meanwhile, Deep Learning is repaying its debt to the life-sciences in some fascinating ways. In the closing Week One interview, Prof. Geoffrey Hinton points to one of them:

"

I guess my main thought is this. If it turns out that back prop[agation] is a really good algorithm for deep learning, then, for sure, evolution could have figured out how to implement it… And I think the brain probably has something that may not be exactly back propagation but is quite close to it.

"

The course will spend a lot of time on the details of "back propagation," but the concept is simple enough. A Deep Learning algorithm looks at its input and guesses the desired answer. Then it compares its guess with the known correct answer (remember, this is supervised/imitable learning.) Finally, it goes back and tweaks the way it came up with its guess and adjusts things to try to guess better next time. This go-back-and-adjust process is the back propagation. Hinton's point is that evolution certainly could have shaped us this way. If it didn't, what did it do instead? Kids have never had this kind of invitation to think about how they themselves think. It is time they did.

Hinton then expands on the underlying approach:

"

There is a huge sea change going on, basically because our relationship to computers has changed. Instead of programming them, we now *show* them. And they figure it out. That's a completely different way of using computers… Showing computers is going to be as big as programming them.

"

Kids still need to learn to program at the same time they learn to read, but not as an end in itself. Computing is way bigger than just programming and getting bigger every year. Tom Vander Ark addresses just this question in his article Skip Coding, Teach Data Science. A resource for going straight to the data is: Data Stories: Data Science For the Middle School Years.

tags: