Which AI project should my students work on for a summer class?

Here’s a request, dear readers. Which AI software should my students work with?

Some context: I’m teaching a summer class on emerging technologies for teaching and learning (and that’s enormously exciting). Right now I plan to have students looking into gaming, 3d printing, and virtual/augmented/mixed reality, plus one or two other, smaller techs still under investigation. Added to all of that is AI.

For each of these technologies students will use the stuff as consumers, read scholarly thinking about it, and also think together through in-person and online conversations. Crucially, I’m also trying to get them making stuff for each. That means learning how to 3d print objects, how to make games (starting with Twine), how to make some basic VR (starting with 360 video), how to make AR, and so on. In short, the class is about both producing and consumption. I’m not pushing them to make publishable, commercial grade products, but to expand their understanding through some making experience, as well as to encourage them to follow up on the tech that intrigues them when the seminar is done.

When it comes to AI, it’s easy to find consumption-side exercises, given the technology’s prevalence in the world. We can look at Google search, computer games managing systems and adversaries, Facebook’s feed, self-driving cars, etc. etc. But I’d like them to enter the production side, even if only for a short while. That might involve training a neural network, tweaking preexisting code, or something else.

There are constraints to this request.

1. The class is very short, just one month long, and you can see that they’ll be very busy with other things. So there’s not a lot of time.
2. I don’t have enrollment data yet, but I’m going to assume a baseline of zero experience with making AI and no experience of coding beyond some HTML work (which isn’t coding per se but you know what I mean). Students are in a graduate program and hence have a wide range of backgrounds; I want to make sure we involve everyone productively, and not just the coders and AI nerds. If we have a decent supply of the latter there will be peer teaching, of course.
3. My support abilities are somewhat constrained. I have little personal experience on the AI production side. I’ve studied the theory and systems a bit, but haven’t run any software. For coding, I did that when I was a teenager, but that last coding I tried my hand with was PHP fifteen years ago. Now, I will have much of May to prepare… but that’s not a lot of time.

There are also strengths. In my experience these students are brilliant. They are keen on making and creativity, and eager to try new things. In addition, the program and Georgetown as a whole offer excellent support.

What would you all recommend, hive mind? Is there a beginner’s neural net package we should try? Did someone launch a generative adversarial network as a service? Does an openly hackable chatbot live online? Or if this is too big a task to attempt, given the constraints, is there an open project my students can participate in, like a machine learning training initiative? Or should we scale back to an algorithm?

Thank you for your thoughts.

(photo by Cogdog; cross-posted to my blog)

Liked it? Take a second to support Bryan Alexander on Patreon!

Related

This entry was posted in teaching. Bookmark the permalink.

Originally published at bryanalexander.org on April 12, 2019.