1.0 Updates:
- none
Ethics in Artificial Intelligence—CSC484
Professor Clark Elliott
Take a walk on the truly wild side with Ethics in AI class!
Waves of novel AI applications are coming to our world. These range from the phenomenally
powerful and useful (self-driving cars) to the outright terrifying (automated fake news). In
this course we will use the algorithmic thinking and logic we are all comfortable with as
computer scientists to explore many aspects of ethical dilemmas facing us as this modern
computational paradigm unfolds around us.
Organization: This course is primarily based on the lectures, the readings and on
extensive online discussion. There may be some light programming / use of AI tools as we become
familiar with the technology. There will be an (algorithmic) writing component as I
occasionally call on you to structure the outline of your arguments in a more formal way. The
class is asynchronous, so you can work ahead by following the study guides and quizzes, but
discussions will follow a progressive schedule. There will be informal Zoom meetings most
weeks as well for those who want to check in as part of our community.
Grading: The focus is on online discussion participation mimicking
an in-class lecture/discussion seminar. There will be: small study
guides/quizzes based on the lectures and readings, and exams on the same;
small written outline assignments; an ethics ideas journal, and a final
ethics paper.
Prerequisites: CSC480 or instructor consent (elliott@depaul.edu) this quarter, where we
are primarily looking for students able to read / listen to lectures and then discuss
sophisticated high-level ideas. I prefer not to exclude students in the Winter 2021 quarter
over prerequisites.
CSC484 is one of the core courses in the new
Artificial Intelligence Degree offered by
DePaul (January 2021) and qualifies as an AI concentration course in CS as well.
Have some fun here with a few interesting ethical problems in
AI:
- Fake News: With automated text
generation, novel faces created from scratch, fake audio added to
altered videos... watch out!
- Cyborg Humans: Computer-brain interfaces
are becoming increasingly sophisticated. Do we implant our children /
soldiers / workers so they can compete in this hyper-evolutionary process?
- The China Brain & Sentience: Human brains are
purely computational devices, except not. Wait... what? Hard problems in ethics
and AI. ←Fun!
- Compassionate Companions: There is a
huge industry push to create artificial companions, in particular for
the elderly. But this rasises intersting ethical choices. For example,
who owns the relationship between a robot and a human?
- Unintentially racially- and gender-biased AI:
Modern AI is really good at capturing existing patterns and generating
novel instances of it, perpetuating existing systemic biases.
- The replication crisis in AI: Because of the intense hardware
demands, many of the leading AI research centers are developed and
supported by industry. The best science requires making algorithms,
platforms and data available for replications, and deep analysis by the
community of scholars. But
industry maintains a proprietary interest in their investments, and also
has a right to industrial secrets. How do we address this challenge?
- AI in the military: Truly autonomous killing machines will
have advantages in the battlefield. If we ever invent true artificial choice, do we
give these devices choice over the life and death of humans? If we
don't, but our enemies do, now what?
- AI for Airport Security and
Lie Detection: An interactive AI Avatar reports 80-85%
acurracy in detecting lies, which is better than most humans (reported
54%). What if
you have "just one of those faces" that the AI Avatar keeps selecting as
suspicious, or your face reflects your chronic
pain? How do you seek redress when the algorithms are secret?
- Face detection: Hoan Ton-That's company Clearview
AI has created an App that let's you upload a picture of someone
and get back photos of that person posted on the Internet, along with
links to the sources. More than 600 law enforcement agencies reportedly
already use this technology. When paired with augmented-reality glasses, this
potentially allows the wearer to get real-time biographies of
strangers on the street. What are the implications for stalkers,
and blackmailers?
- Human memory transfer via electronic means may
not be as far off as we think. What are the implications for privay
and individuality?
- Artificial Sentience: What do we do if
we stumble upon artificial life and sentience? This opens Pandora's box
in the realm of ethical behavior for humans in the (coming?) future.
- Basic Ethics: A flash visit with two
ethical frameworks: Utilitarianism and Kant.