About This Workshop
Xerox Research Scientist, Adrien Gaidon, recently remarked that, “It’s hard for a human to teach AI because they are worse than the worst toddlers in the world—you have to explain everything.” It might be hard for human beings to teach AI, but video games like Atari, Go, poker, Jeopardy, Super Mario World, and Minecraft (among others), are doing an astounding job of teaching algorithms (the underpinnings of Artificial Intelligence) to learn how to "understand" and navigate the complicated world we inhabit. But how are videos able to train AI so well and what are the implications for the development of AI?
JOIN US for this special, interactive workshop featuring guest instructor, Julian Togelius (Assoc. Professor, Artificial Intelligence in Games, NYU Tandon School of Engineering), as we delve into the fascinating world of training algorithms using various types of games (from board games, to video games, to strategic games).
What Will This Workshop Cover?
Dr. Togelius will give us a brief but insightful overview of the history of gaming being used to train AI, highlight some of the most notable and shocking advancements in AI that have come about as a result of using games to train them (ie, Alpha Go, Jeopardy, etc.), and get us up to speed on the latest research and findings in this space.
Questions and topics we'll be reviewing (among other things) include:
- What exactly are we trying to train algorithms to do using games? What are the expectations of researchers working in this space?
- How are games helping us to define Artificial General Intelligence?
- What are some of the challenges and disappointments researchers have experienced training AI using games?
- What are the limitations of teaching AI using games?
- How are AI consortiums like Elon Musk's OpenAI using games to train AI and what are their goals?
- How are games being used to teach driverless car algorithms to drive?
- How does the research and development in this space impact businesses (from startups, to enterprise)?
- As AI improves surprisingly quickly, and literally beats us at our own games, what are we learning about ourselves that we didn't know before training AI using games?
- Are there games that can teach human beings about how to understand and work with AI (ie, robots)?
After the presentation, we will have our popular interactive, group experience where you will get the opportunity to explore and answer some of the challenging problems in this space with others who are just as intrigued by this topic as you are!
No technical experience required. This is a non-technical workshop meant for everyone: marketers, advertisers, developers, engineers, product managers, investors, data analysts, students, policy wonks -- all are welcome! Here's a few articles on recent developments in this space that you may find informative and helpful:
- Google is using old Atari games to teach artificial intelligence how to learn
- Elon Musk’s Group Wants to Use Video Games to Teach AI About Life
- Making AI Play Lots of Video Games Could Be Huge
- Why AI Systems Are Learning to Play Old-School Video Games
- Don't Worry, Driverless Cars Are Learning From Grand Theft Auto
- In a Huge Breakthrough, Googles AI Beats a Top Player at the Game of Go
- Google Wins on Jeopardy: Trivial, It's Not
- IBM Watson: The Inside Story of How the Jeopardy-Winning Supercomputer Was Born, and What It Wants to Do Next
- Artificial Intelligence Goes Deep to Beat Humans at Poker
- How To Watch the Brains Vs. AI Poker Rematch
- Kick back and watch our AI learn to play Atari games in real-time
- 6pm - 6:15pm - Sign in, enjoy lite bites and beverages
- 6:15 - 6:20 - Introduction, Tech 2025 announcements, sponsor acknowledgement
- 6:20 - 6:45 - Guest instructor, Julian Togelius, presentation
- 6:40 - 7:50 - Interactive group exercises and problem-solving
- 7:50 - 8pm - Final thoughts and Q&A
Julian Togelius, Associate Professor, Artificial Intelligence in Games, Procedural Content Generation, Player Modelling, Games Researcher, NYU Tandon School of Engineering.
I work on artificial intelligence and games. I research and develop methods for making games more fun, easier to design and develop, more adaptive, or just to enable games and interactive experiences that we cannot yet create. I'm interested in all kinds of games that people actually play: video games, board games, card games or mind games.
When it comes to AI techniques I'm flexible and willing to engage with all kinds of methods, but my methodological roots are in evolutionary computation and neural networks.
Some keywords that describe what I do are game AI, player modelling, procedural content generation, automatic game design, believable bot behaviour, coevolution, neuroevolution, genetic programming and monte carlo tree search.
Our Awesome Workshop Sponsor
In 2017, Galvanize will be bi-coastal and New York City will mark our ninth campus. Technology is among the fastest growing and highest paying sectors in New York and the Galvanize NYC campus is poised to help meet the growing tech talent demand. Galvanize New York is located in vibrant Lower Manhattan. Check out our events, tour our campus, be inspired to learn, and build your company as we prepare for the grand opening. Contact us at: firstname.lastname@example.org / Twitter: @galvanize / Facebook: facebook.com/pg/galvanizeNYC