The main objective of the rest of our time together is to understand how, at a high level, how AI models can generate dialogues, code, images, and other kinds of data.
We’ll also mention some other topics that you should be aware of, but that we won’t have time to cover in detail.
(By “work”, I mean, how do they use the building blocks of neural networks, which we’ve studied so far, to generate data.)
Here are a few more things that would be helpful to know about, even though they’re not the main focus of this class:
Please read the following sometime before the last day of class:
Review my blog post on Mapping to Mimicry. I wrote it in one short sprint; feedback welcome!
The Illustrated Stable Diffusion – Jay Alammar – Visualizing machine learning one concept at a time.
Rather than study theory, let’s look at two recent blog advances:
What happens when AI meets people? How can we ensure that AI results are:
The first two are the subject of a subfield called Fairness, Accountability, and Transparency; the last is the subject of much research in human-computer interaction (HCI) and computer-supported cooperative work (CSCW). We’ll explore all three in these last two weeks of class.
Read one or more of these:
Many unwittingly used a data set that contained chest scans of children who did not have covid as their examples of what non-covid cases looked like. But as a result, the AIs learned to identify kids, not covid.
Watch:
Supplemental: The Effects of Regularization and Data Augmentation are Class Dependent | Abstract
Read or watch something from Human-Centered Artificial Intelligence.
No Class Monday (Easter Monday)
Generating images
Lab: generation: autoregressive, GAN, diffusion.
Interpretability and Explanation (slides)
Fairness and Wrap-Up slides
Final Discussion topics