An Open Letter to Udacity CEO Sebastian Thrun:

With tremendous gratitude for bringing us together to learn some of the most important emerging skills and technologies of our time, we your passionate students present to you a petition for an interesting second (or third?) term project for AIND. These signatures were gathered over the course of only 12 hours! When we heard the proposal from our fellow student Adrian Melic, we thought it was too good to be ignored, and I'm sure you will immediately see the value when you take a look at it!

Although the proposal speaks for itself, the impetus behind it is in reaction to the choices for the second term project. The options presented during the AMA-session were CV, NLP, and speech recognition. While each is respectively a fascinating and worthy topic, we aspire to use those tools in conjunction with others to accomplish something extraordinary -- something fun and easily appreciated that brings together both experienced industry professionals and the most non-technical among us who we support with our efforts -- something truly worthy of the lofty title "Artificial Intelligence".

Both a means to making many powerful techniques accessible and an enjoyable end unto itself, and the potential basis for a fantastic Open Source Udacity General Game Player Artifical Intelligence, we present to you the proposal for the Game Agent AI Project:

@adrianmelic:

Speech Recognition seems almost "solved". Amazon, Google, IBM, Baidu... Really great interpreters.

*What if Speech Recognition is changed to "Game Agent AI"?*

After Google DeepMind's victory at Go, in the beginning of 2017 they will release StarCraft II as an AI research environment. OpenAI has a gym where you fit an agent playing games. Microsoft with project Malmo (Minecraft AI agent). Facebook with Doom... Everything we can create in a game, a simulation, can be tried later in the real world. Every business is just some kind of game to be optimized. Maybe in the future we create a micro-version of The Matrix :sunglasses: but right now we need some great course that includes this. It will be super fun to watch shared videos of Udacity's student AI agents playing, competing, trying to help OpenAI or winning to a human.

Specifically, we would like to see the use of model-free reinforcement learning such as Q-Learning in conjunction with these games as a medium to help us achieve fantastic ends! Some comments by your students:

"I am writing an email to OpenAI CEO. It will be so cool if we help them democratize AI. Maybe some of our code can be in a Mars base builder robot when SpaceX launch the firsts missions to Mars" ~ @adrianmelic

"If they included reinforcement learning for games as a term 2 focus I would take it 12 times" ~ @nash

"That could be a nice enough challenge to have a bot flexible enough to handle the difference of environment while still staying in a context that is similar to the one it was trained with" - @kavignon

"that would be great if we could work on something related to reinforcement/q learning and OpenAI Gym" ~@jro

"I am also very interested in "Game Agent AI", especially if more "game theory" content is included. Game theory will be a crucial knowledge to determine/prevent the robot before it go against us.... I will take it too. Count me in." ~@kora

"I definitely see a lot more value add from a gameAI concentration than speech recognition. Decision support in the future will be infinitely more valuable than speech recognition -- being able to take confident, autonomous actions produces so much more business value across industries than speech recognition which has diminishing research returns at this point" ~ @mattzhou

We need all the help we can get. It will be very difficult to adopt this, as pointed out:

"Does the Udacity team even 6 pool?" ~ @madhavajay

(hint: https://www.google.com/search?q=zerg+rush)

Gauntlet has been thrown, Udacity! ;)

Don't worry! We understand this may be a daunting task and we are 100% committed to both helping develop and execute the new curriculum.

Sincerely,

The undersigned.