6.xxx (AKA 6.803 and 6.833)

The Human Intelligence Enterprise: 2017

Home
Prospectus
FAQ
Staff
Schedule
Style Guide
Grading Marks
Project Instructions
Office hours

Papers

Read the assignment before you read the papers.

ImageNet Classification with Deep Convolutional Neural Nets, by Alex Krizhevsky, Illya Sutskever, and Geoffrey E. Hinton.

WRT: 20 min

Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images, by Anh Nguyen, Jason Yosinsky, and Jeff Clune

WRT: 30 min

Mastering the game of Go with deep neural networks and tree search, by David Silver et al.

WRT: 30 min (not including optional Methods section)

Note that W read these in big picture mode, not problem set mode. Note that you are only expected to read the first and skim the others.

Note that much of the Hinton paper, in particular, is unintelligeable.

Assignment

If you discuss the paper or the assignment with another student—which we encourage—indicate whom you have talked with in your submitted composition. Of course your submitted composition must be written entirely by you.

On a total of one side of one sheet of paper, using 10 pt type or larger, with standard interline spacing and margins, respond to all the following:

Imagine that Winston, having grown tired of teaching, has left MIT and started an international consulting firm, The Concord Group, specializing in giving advice about AI. He has hired you as an intern because you took 6xxx.

On your first day, he says, “Good news. We've just been hired by Japan's Ministry of International Trade and Industry, also known as MITI. The MITI people view current trends as the third wave of AI.

They wonder if deep neural nets are The Answer and whether they should start a big research program in Japan to do research on deep neural nets. They need a white paper to help them decide, and you get to write it!”

First, figure out what a white paper is.

Then, at Winston's suggestion, you decide to base your white paper on Hinton's paper, so you read that. Then, you skim the paper by Nguyen et al., looking only at the pictures and reading the captions. Finally you skim the AlphaGo paper, hoping you can learn why success in image classification contributes to success in playing Go.