What We Talk About When We Talk About AI
Library Futures was pleased to welcome Dr. Chris Gilliard and Dr. Christa Albrecht-Crane for “What We Talk About When We Talk About AI,” the first installment of our fall series Machine Learning and Artificial Intelligence for Information Professionals.
In a talk that posed provocative questions from scholars and practitioners of machine learning, Dr. Gilliard encouraged the audience to consider the ways in which AI is fundamentally a surveillance technology and a political artifact, not a technological one. The audience for AI is not the public: it is governments, investors, corporations, and those who seek to surveille and control. According to Dr. Gilliard, the people who build these systems seek to create a kind of robot god. When we see AI as its makers wish us to, we fall for that power masquerading as knowledge.
Dr. Gilliard typically speaks with his camera off and does not record his talks. Why? As he noted to the audience, artificial systems have invaded our online spaces, ingesting anything they can find and turning it into fodder for large language models. “Don’t feed the machine,” he said, noting that his decisions about cameras and recording stem from a desire to resist that machine whenever possible.
Dr. Gilliard encouraged anyone interested in pursuing these concepts to check out the work of the scholars who guided his thinking, including: Timnit Gebru, Melanie Mitchell, Emily Bender & Alex Hanna, Abeba Birhane, Karen Hao, Iris van Rooij, Olivia Guest, Eryk Salvaggio, and Dan McQuillan.
Dr. Christa Albrecht-Crane found similar inspiration in her research on “AI” and large language models. Her presentation focused on pulling back the curtain that often obscures the mathematical functioning of LLMs to “dispel the illusion” of a magical and all-knowing interlocutor. To demonstrate, Dr. Albrecht-Crane took the audience on a tour of NovelAI, a “raw” LLM advertised for anime art and stories. In the video below, she shows how the platform translates text into numerical tokens and how it offers textual suggestions based on statistical calculations. Viewers should be warned that the system responds predictably, offering up a story of violence against women and revealing the underlying bias of the dataset used to train the model.
Links and Further Reading Suggestions
From Dr. Chris Gilliard
Dr. Gilliard referred to the following sources during his presentation and Q&A:
- Is AI really trying to escape human control and blackmail people?
- Ali Alkhatib: Defining AI
- The AI Con
- Any Sufficiently Transparent Magic…1
- The New Artificial Intelligentsia
- Weapons of Math Destruction
- Weizenbaum examines computers and society
From Dr. Christa Albrecht-Crane
Dr. Albrecht-Crane referred to the following sources during her presentation and Q&A:
- The semantic illusion under the hood of LLMs
- Thinking Smarter, not Harder? Google NotebookLM’s Misalignment Problem in Education
- Who are we talking to when we talk to these bots?
- Why You’re Thinking About “Reasoning” All Wrong
- From HAL to Help: AI’s Role in Learning Assistance
From the Chat
In addition to the resources shared by Drs. Gilliard and Albrecht-Crane, webinar attendees shared a wealth of links, which we’ve gathered below.
Books
- More Everything Forever by Adam Becker
- Your Face Belongs to Us by Kashmir Hill
Articles and Talks
- Neon, the No. 2 social app on the Apple App Store, pays users to record their phone calls and sells data to AI firms
- The State of Personal Online Security and Confidentiality
- Against the Uncritical Adoption of ‘AI’ Technologies in Academia
- Investigating the “Feeling Rules” of Generative AI and Imagining Alternative Futures
- Word2Vec Explained: Explaining the Intuition of Word2Vec & Implementing it in Python
- Studying Up Machine Learning Data: Why Talk About Bias When We Mean Power?
- Generative ML and CSAM: Implications and Mitigations
- AI-Generated “Workslop” Is Destroying Productivity
- Sometimes We Resist AI for Good Reasons
- I’m a High Schooler. AI Is Demolishing My Education
Tools for Teaching
Several people mentioned other sources like NovelAI that could be useful in helping students understand the mechanics of large language models and machine learning.
- Going beyond open data – increasing transparency and trust in language models with OLMoTrace
- Semantle
- Contexto
Conferences
A couple of upcoming conferences of possible interest to attendees.
