Artificial Intelligence and Machine Learning
Computers are getting progressively more interactive and present in everyday life.
In your group select one of the applications below that was not demonstrated in class. Take one or more screen shots to capture what the application does, and as a group answer the following questions
How do you think it works? Where does the data come from that it uses?
How does it use data for pattern matching or inferencing?
How might this be extended and/or used in some meaningful context?
How might it be helpful to people, or how might it be destructive?
Applications
The four demos in the set below are from Google's AI Experiments page:
Quick, Draw! recognizes what you sketch.
Based off of accumulated drawings, sketch-rnn takes what you start and completes the sketch, and autodraw.com recognizes what you are drawing and gives you versions to choose from.
Google's Blob Opera allows you to drag up and down for pitch, forward and back for vowel sounds, and selecting which blob to control. This was trained with Machine Learning, starting with real voices.
For a picture or Street View, listen to the estimated accompanying sound in Imaginary Soundscape. (Experiment with the "Random" button at upper-right). You can also supply a link to an online picture to be analyzed.
See how people have associated words using the Sematris games.
Similarly take a picture and have thing-translator recognize it and translate into some other language.
See if you can tell the difference between poetry written by humans or by computers at http://botpoet.com/quiz/dwf/
Others:
Consider an algorithm that matches patterns to previously seen images and collaboratively creates, such as in True Colors.
This is similar to the pattern matching used on brain waves (after training using an fMRI machine) that lets you "see" what a person is thinking about.
Click on portions of one trained brain model to show which type of words are mapped to which brain areas.
Microsoft's https://www.how-old.net/ (may be slow) guesses the age of people in a photo.
Face recognition is used in many different applications, but Joy Buolamwini points out the perils of over dependence.
At bit.ly/aiguess see if you can out-guess a computer doing pattern matching, that tries to forecast if your next guess will be 0 or 1. Select "Show Data" to see the data tables behind the scenes to see how it learns.
While not interactive, Deep Fakes apply Machine Learning to combine audio and visuals to create artificial characters that look and sound like the real thing, only they're not. This is often done with/to public figures. For example, see: Obama/Jordan Peele, or the bus scene from the movie Forrest Gump, where you can compare the original and the Keanu Reeves version. See https://thispersondoesnotexist.com and scroll through the faces in this NYT article.