becoming aware with how little control we have...
and how much grace we need -- Thomas Merton
If there is one upside to the constant drone of AI this and AI that, I have to say it is the focus on the difference between conscious and unconscious cognition. My origin as a bench scientist always has me looking for the scientists in the room. It reminds me how Fred Rogers shared that his mother always told him to “look for the helpers” during disasters. They are the beacon of hope needed to see a way forward.
Perhaps not as flashy as sensational headlines and ledes, but facts matter. We rely on satellite data and the daily petabytes infusion of information but what does it all mean without our conscious attention? Location intelligence is one part of the puzzle but unless we engage with “place” we do not have story building capabilities. Now that everyone is a storyteller (cough, cough) it is time to move beyond the “where” something is happening to the actual boots on the ground questions of moving about the place, considering culture, accessibility — what it is like to live and work there.
Although it isn’t possible to capture every intellectual crossroad, the ability to think critically, even scientifically if you will — is a practice. Early days of blogging were primarily a way for me to catalog interesting stories, videos and articles. My hopes are to move the archive into a print book but for now you can dig around data & donuts.
I had a wee accident over the weekend. I slipped down a few hard wooden stairs. In an abundance of caution workouts and running are lower volume. The upside is the ability to dig into books and items on the watch list. There a few things that I can’t stop thinking about. Somehow they will find their way into a busy fall of keynotes and invited talks. I wanted to share them with you as fodder for future conversations.
Let’s start with the most engaging short explanation of concept cells. These are cells associated with a memory or an idea. A way of cataloging experiences from our daily lives. Your lived experiences inform your intellectual framework. It seems the concept cells are there for all of it.
Picture the opening scene described in the podcast. Sitting in a bar waiting for your date to arrive. There is the location (have you been to this bar before?), the drink, and a whole host of multiple single neurons firing to create a multi-faced memory.
The reliance on these cells to create memory functions and flow of consciousness are unique to the human experience. Think about how this limits the processing possible in LLMs. They lack experiential depth although maybe we aren’t as “deep” as we think we are…
Somewhat related in concept is the AI Interview from the American Voices Project. The quick take is an AI prompt asks a series of conversational open-ended questions. After the data is collected, the AI is instructed to answer a series of questions based on your persona gleaned through the responses of the in depth survey. You are presented with the same questions looking for concordance between you and the AI local LLM.
Questions vary from the first, I would like to begin with a big question:
Tell me the story of your life. Start from the beginning—from your childhood, moving on to Some people tell us that they've reached a crossroads at some points in their life where multiple paths were available, and their choice then made a significant difference in defining who they are.
What about you? Was there a moment like that for you, and if so, could you tell me the whole story about that from start to finish?, How have you been thinking about race in the U.S. recently?, Tell me a story about a time in the last year when you were in a rough place or struggling emotionally and more.
The American Voices Project (AVP) interview script contains 118 main questions across a wide range of topics. Here's a categorized summary of the question topics asked:
Introduction
Name, age, and gender
Life Story and Turning Points
Life story (childhood, education, family, major events)
2–4. Life crossroads and decision-making
Family and Relationships
5–7. Important family members and relationships outside family
Neighborhood and Living Situation
8–14. Description of neighborhood and household members
Daily Life and Routines
15–17. Daily routine, work schedule, and responsibilities
Law Enforcement and Political Views
18–23. Experiences with law enforcement and political orientation
Race, Racism, and Policing
24–29. Thoughts and responses to race and racism issues, including BLM
Health and Healthcare
30–42. Health status, challenges, healthcare access and coping mechanisms
Emotional and Mental Health
43–46. Emotional state over the past year, struggles, and family wellbeing
Religion and Social Media
47–50. Importance of religion, use of social media, and online support
Emotional Coping and Living Situation Recap
51–56. Emotional coping, household composition, shared responsibilities
Finances and Employment
57–75. Budgeting, expenses, housing, savings, debts, income, and taxes
Workplace Environment
76–82. Workplace tenure, benefits, relationships, flexibility, and childcare
Public Assistance and Financial Shocks
83–86. Use of public assistance, financial instability, emergency expenses
Marital Status and Demographics
87–96. Marital/cohabitation status, birth details, nationality, education
Military Service and Religion
97–98. Military service, religious identity
Political Party Affiliation
99–101. Political identity and strength of affiliation
Parental Background
102–116. Residence at 16, parents' education and work history
Hopes and Values
117–118. Hopes for the future and personal values
Lots to discuss here but the goal is to go after that secret sauce, “... activating the data system through qualitative data in large language models so this rests on new and still emerging evidence that qualitative data and in particular American voices project data can provide dark matter infused behavioral blueprints ...” Professor David B Grusky
Again, acknowledgment that we are coming up short when “humanizing” AI. For the record, the part that intrigues me is the attempt to find contextual depth beyond administrative survey responses. Not to give away the plot but the concordance is pretty high…
Your mileage may vary but I would love to discuss this in more detail…
In the role of the viola in the symphony (brings the harmony and melody together) is this book that I simultaneously began reading while all these other ideas were swimming around my head.
Yet our increasing reliance on computational media means that decisions, like agency, are distributed throughout the collectivities of humans, nonhumans, and computational media…Even when humans appear to be in control, their assumptions have been formed and mediated by prior decisions and interpretations made by computational media, so that they can scarcely be considered autonomous or self-determining.—N. Katherine Hayles
“In waiting, we become aware with how little control we have — and how much grace we need.” — Thomas Merton