- [] /
- 09 Jan 2020
- Artificial Intelligence Machine Learning Complex Systems
Complexity Explorer January Newsletter
Wishing the Complexity Explorer community a happy New Year from frosty Santa Fe, NM. 2020 is shaping up to be another great year at the Santa Fe Institute. First up we have two courses starting next week - see info below - and many exciting things planned. At the end of 2019, we hosted Melanie Mitchell in Santa Fe for a lecture about her new book on Artificial Intelligence. We got to ask her a few questions about the book and her process and are excited to share the interview with you all.
Interview with Melanie Mitchell
You started Complexity Explorer nearly a decade ago, and your first “Introduction to Complexity” course launched in 2013. At roughly the same time, you published your hugely popular book “Complexity: A Guided Tour”. Why did you take on all of these projects at that time?
Actually, my book, “Complexity: A Guided Tour”, was completed in 2008 and published in 2009. I started working on the idea for Complexity Explorer (partnering with Ginger Richardson, former SFI VP for Education) a year or so later. There were a few reasons I wanted to launch an online course platform at SFI. Many people really wanted to come to SFI’s summer school or one of our short courses, but were unable to, due to financial or other logistical reasons, and also because we simply didn’t have enough space! Ginger and I were looking for a creative way to enable people to be involved with SFI, to learn about complex systems, and were talking a lot about putting short courses online. Around this time Stanford started offering some of the first Massive Open Online Courses (MOOCs) and they were wildly successful, so this pushed us to think about offering longer-term, focused courses. I had also been working on turning my book into a course at my own university, so it seemed obvious that I should offer this course as the first SFI MOOC. We managed to obtain funding from the Templeton Foundation, and three years later we were ready to launch this MOOC!
We still get thousands of people registering for “Introduction to Complexity”. What do you hope they take away from the course?
There are several ways to “take” this course. You can go through all the units in order, doing all the exercises and homework, just as one would do in a traditional course. Or else you can skip around, watching videos of particular interest, doing some or even none of the homework. Obviously, the more material you cover, the more you will get out of the course, but I know that people have different goals. My main hope is that people gain some new insight into how the world works, and get a sense of why complex systems researchers are so excited about what they do.
What was your biggest motivation for writing a book on Artificial Intelligence? Are there particular issues in AI that you find especially resonant?I found that I was confused myself about the current state of AI, and wanted to understand what the true state of the art was, and how close we actually are to “human-level” or “general” AI. In the book, I talk about a particular meeting I attended at Google in 2013—the meeting featured both incredible optimism and deep “terror” about the prospects for general AI. It was after this meeting that I got the idea for writing a book to share my own explorations into several areas of modern AI, as well as into some central issues in human intelligence. It “only” took five more years for the book to be done!
What was the most surprising thing you learned in researching your book?
I continue to be surprised at how statistical learning methods, including deep neural networks, can do so well in so many areas without actually understanding their input in any humanlike way. One example is speech recognition. You can dictate texts and emails to your phone and it will transcribe your speech almost perfectly, but it doesn’t understand the meaning of what you are saying. Same with automated translation: you can type in a sentence in English and it will be translated instantly into, say, Chinese. The machine can do a very good job without understanding the meaning of the input or the output. I would have predicted that this is not possible, but statistics over huge datasets have turned out to be very powerful. Of course, my statement requires me to explain what I mean by “understanding” or “meaning” here. I spend several chapters of my book talking about exactly that issue, which I see as the fundamental one for AI.
MORE FROM MELANIE
Have you had a chance to watch the lecture Melanie gave at SFI in November?
MORE ON MACHINE LEARNING
Machine Learning is a fast growing, rapidly advancing field that touches nearly everyone's lives. There has recently been an explosion of successful machine learning applications - in everything from voice recognition to text analysis to deeper insights for researchers. While common and frequently talked about, most people have only a vague concept of how machine learning actually works.SFI postdoctoral researchers Dr. Artemy Kolchinsky and Dr. Brendan Tracey created a tutorial to outline exactly what it is that makes machine learning so special in an accessible way. The principles of training and generalization in machine learning are explained with ample metaphors and visual intuitions, an extended analysis of machine learning in games provides a thorough example, and a closer look at the deep neural nets that are the core of successful machine learning. Finally, it addresses when it's appropriate to use (and not use) machine learning in problem solving, as well as an example of scientific research incorporating machine learning principles.
UPCOMING COURSES
Introduction to Agent-Based Modeling
Launches on January 15th
Scholarships Available
Nonlinear Dynamics: Mathematical and Computational Approaches
Launches on January 15th
