Schedule

GUIDING QUESTIONS:

What is AI? How does it work?

BRIEF DESCRIPTION:

Today focuses on defining AI, understanding how AI works, and exploring some prominent AI failures.

GUEST:

Rogelio Cardona Rivera (University of Utah, Division of Games)

READINGS:

• Pre-institute book: You Look Like a Thing and I Love You, Janelle Shane.

MORNING (9:00-12:00):

• Participant introductions
• Collaboratively come up with our AI policy and our principles of civility for the institute
• Definitions of AI and the slipperiness of the term.
• Typology of AI.
• Prominent AI failures.
• Collectively share our experiences with AI in teaching and research
• Introduce institute projects (podcast, lesson plan, exhibit).

AFTERNOON (1:00-4:00):

• Code-along with line-by-line explanation in Google Co-labs of an AI system that recognizes handwritten numbers.

EVENING:

Welcome Dinner

Details TBA

GUIDING QUESTIONS:

What is a large language model? How does it write? How is its method of writing different from human writing? What consequences stem from this method of writing?

BRIEF DESCRIPTION:

Today’s seminar focuses on one particular type of AI, the large language model (LLM). We will explore LLMs, their inner workings, their science fiction predecessors, and the repercussions of their release.

GUEST:

Lisa Medeen (Swarthmore College, Computer Science)

READINGS:

• “What is an LLM?” Elizabeth Callaway.
• “EPICAC,” Kurt Vonnegut
• Selection of all-time greatest love poems.
• “Chat GPT Hallucinates. Some researchers worry it isn’t fixable,” Washington Post.
• “Synthetic Humanity: AI & What’s At Stake” Podcast episode, Your Undivided Attention.

MORNING (9:00-12:00)

• Guest lecture on LLMs and ChatGPT.
• Workshop coding a “small language model”

AFTERNOON (1:00-4:00):

• Play Wise and Otherwise with ChatGPT as a player
• Have ChatGPT write a love poem. Discussion comparing to Vonnegut story and love poem packet leading into discussion about skills ChatGPT is good at and skills it isn’t competent at yet. 
• Tour ARPA node collection, Marriott Library.

GUIDING QUESTIONS:

Where does AI data come from? How can data be unfair and why is it so?

BRIEF DESCRIPTION:

Today we dig into the all-important question of data. We learn what data powers AI and begin to explore the ways that data can be corrupt.

VIRTUAL GUEST:

Lauren Klein (Emory University, Quantitative Theory & Methods and English)

READINGS:

• From Data Feminism “Introduction” and “What Gets Counted Counts”
• “The Hidden Biases in Big Data,” MIT Technology Review.
• “Introduction” and “Data Before the Fact,” from “Raw Data” is an Oxymoron.
• ““Inferences & Connections,” Social Science Research Network. 

MORNING (9:00-12:00)

• Virtual Guest lecture 
• Discussion on data, bias, and AI

AFTERNOON (1:00-4:00):

• Explore a dataset and find the bias. We will have sample datasets, or you can bring your own.
• Read together “Data Biographies: Getting to Know Your Data” and create data biographies in groups.

GUIDING QUESTIONS:

What does responsible AI look like? How is Adobe engaging with AI?

BRIEF DESCRIPTION:

Today we will hear from Adobe, an industry leader in responsible AI. We will tour their silicon slopes headquarters on a field trip to Lehi, Utah.

GUEST:

Brian Johnsrud (Adobe Inc., Director of Education, Learning, and Advocacy)

READINGS:

• None

MORNING (9:00-12:00)

• Bus to Adobe
• Tour Adobe

AFTERNOON (12:00-4:00):

• Lunch together
• Adobe demonstrations
• Bus back to University of Utah

GUIDING QUESTIONS:

How do you make a podcast? How does one use podcasting in teaching?

BRIEF DESCRIPTION:

We will devote today to making a podcast and finalizing our content for the institute exhibit.

GUEST:

Matthew Winters (Utah State Board of Education, AI Specialist)

READINGS:

• Whatever is required for your podcast subject.

MORNING (9:00-12:00):

• Workshop on how to make a podcast.

AFTERNOON (1:00-4:00):

• Start recording podcast.
• Learn how to edit podcast.

GUIDING QUESTIONS:

How do AI incentives manifest in environmental impact? What are the carbon and water footprints of AI? Why is it so hard to know exactly?  What are the connections between material extraction (mines) and data extraction?

BRIEF DESCRIPTION:

Today takes as its focus the environmental costs of AI. Often using AI tools feels identical to using any software. Using a Microsoft Word document to write and using ChatGPT to write feel identical, but they have vastly different carbon and water footprints. However, the complexity and lack of transparency and accounting make it difficult to even determine the environmental impact of AI.

GUEST:

Raphaël Deberdt (Colorado School of Mines, Department of Mining Engineering)

READINGS:

• Anatomy of an AI System: The Amazon Echo as an anatomical map of human labor, data and planetary resources by Kate Crawford.
• “AI can help us fight climate change, but it has an energy problem too,” European Commission. 
• “You should worry about the data retailers collect about you.” The Atlantic.
• Play Universal Paperclips

MORNING (9:00-12:00):

• Guest Lecture and Discussion.
• Jigsaw discussion on types of extraction.

AFTERNOON (12:00-4:00):

• Lunch and field trip to the Caldecott Copper Mine.
• The role of copper in AI computation.

GUIDING QUESTIONS:

How much is AI just a condensation and remixing human work? What changes if we view AI as the product of human labor instead of machine intelligence?  How do we design AI that extends human capabilities rather than replaces humans?

BRIEF DESCRIPTION:

Today we see the underbelly of AI development through readings about the nitty-gritty of “reinforcement learning with human feedback,” a process that is fundamental to creating AIs that actually sound human. There is a mountain of human labor behind AI that goes unacknowledged and is often exploitative and unfair to workers in the global South.

GUEST:

None

READINGS:

• Inside the AI Factory, The Verge.
• The Turing Trap: Problems and Perils of Humanlike AI, Stanford.
• Excerpts from Of the Subcontract.
• “Spew,” Neal Stephenson.
• “OpenAI’s hunger for data is coming back to bite it,” MIT Technology Review.
• “Origin Stories: Plantations and Computers,” Logic(s).
• Excerpts from We Had to Remove This Post.

MORNING (9:00-12:00):

• Working conditions in the global South.
• Difference between AI labor and sweatshop labor.

AFTERNOON (12:00-4:00):

• Play “I agree to the terms”
• Play Moderator Mayhem.
• Why are labor games an interesting medium for exploring labor in AI?
• Plan final projects

GUIDING QUESTIONS:

What are the consequences of our current misinformation crisis?

BRIEF DESCRIPTION:

Today we explore the state of mis-and-disinformation and the role that social media AI plays in it. We also think through some of the potential changes generative AI is poised to make to the information environment.

GUEST:

Claire Wardle (Information Futures Lab & Cornell University)

READINGS:

• “Stranger than Fiction,” Your Undivided Attention by Claire Wardle.
• “A Genocide Incited on Facebook, With Posts from Myanmar’s Military” The New York Times.
• “Bot Holiday: Covid disinformation down as social media pivot to Ukraine,” Melodie Schrieber.
• “Birds Aren’t Real? How a Conspiracy Takes Flight,” TED Talk.
• “Teen Girls Confront an Epidemic of Deep Fake Nudes in Schools,” The New York Times.
• Chapter 1: The Problem of Misinformation and Disinformation from Misinformation and Disinformation: Detecting Fakes with the Eye and AI by Victoria L. Rubin

MORNING (9:00-12:00):

• Guest Lecture and Discussion, Misinformation and Censorship.
• Explore Solutions.

AFTERNOON (1:00-4:00):

• Field Trip to the Salt Lake Tribune with Communications Professor Avery Holton

GUIDING QUESTIONS:

What are the causes of the misinformation crisis? 

BRIEF DESCRIPTION:

We will read about the attention economy and how it bolsters and amplifies misinformation while we search the past for solutions to previous information crises.

GUEST:

None

READINGS:

• Down the Rabbit Hole by Design, Your Undivided Attention Gaulliome Chaslot.
• “A glimpse into facebook’s opaque and potentially vulnerable trending algorithm.” Dave Gershgorn, Mike Murphy. Quartz, 2016.
• Fueling the Fire: How Social Media Intensifies Political Polarization. From The Attention Merchants.

MORNING (9:00-12:00):

• Attention economy.
• Conspiracy theories.
• Radicalization

AFTERNOON (1:00-4:00):

• Pioneer Day Party @ Rebekah’s house

GUIDING QUESTIONS:

What have we learned? How will we implement what we’ve learned into our research and teaching?

BRIEF DESCRIPTION:

Today we reflect and celebrate.

GUEST:

None

READINGS:

• None

MORNING (9:00-12:00):

• Panel on Humanities and responsible AI
• Reflection on what we learned. What was most useful? What do you wish you learned that you didn’t. 
• Make a plan for implementation of institute knowledge in teaching and research.
• Brainstorm white paper. 

AFTERNOON (1:00-4:00):

• Share our podcasts
• Final celebration.