MyAI – Gaining Insight Into Your Own Data
MyAI – Gaining Insight Into Your Own Data
Notes-taker(s): Travis Giggy
Discussion notes, key understandings, outstanding questions, observations, and, if appropriate to this discussion: action items, next steps:
Notes for session: My AI
Facilitated by Sari: firstname.lastname@example.org Emailed to: email@example.com
What do we do with our data?
Starting with requests for topics:
Frequent flyer miles
- How can we shape a new loyalty? How do we re-make loyalty?
- Our own data has "use value" but little value in the marketplace.
A better relationship with United Airlines than now.
A better relationship with himself
- The basement of my life is so full I can't get downstairs
So much potential. What are the first steps we can be taking to give ourselves more insight to ourselves?
An idea to make a bathroom scale w/o a number readout. You have an AI that will give general advice about life. No angst about a number, just
"Jobs to be done list"
Travis: Organize your life and achieve your goals
How will he know that the AI will be controlled by him? How does he know that it has his best interests in mind?
Travis: Trust and companionship
Another way: do the computing at the edge on your own personal devices
Think of it as digital army.
Digital mirror. Lifelock on crack
Something that spins up bots and manages needs
Big companies have big data and analyze you. Why can't I do that for myself?
Sari wants to break up into groups
Instead, by suggestion, we are going to have all the people in the room who are working on something similar talk about their viewpoint for a couple of minutes each.
Sam Chase: Lifescope.io
- Builds mesh networks using webER in the browser, raspberry pi's
- Lifescope is a master API collector, grabs data bout you, collects it, you can learn about yourself. Trying to create the bedrock of a personal AI marketplace. Trying to create perfect context, allowing you to have streams *into your life, more real-time AI.
- They have a wallet working
- Right now, OAuth, GraphQL, they are here for people to look at what they're doing, how to make it more secure, scalable, something that people want to use.
Rob Collins: TodayPlatform.com
- UI layer
- We have gone from local villages, the EU has an idea... if you want to shop for shoes, there are 100k-1M places to shop. If you go to Google and look at privacy policies, it would take 4 years to buy shoes. We need to speed that process up and extends our brain to be faster.
- We don't need AI, just ordinary logic
- A public benefit corp, measures success by the formula: Number of humans * Depth of relationship * Increase in well-being
- Creating a consortium of large enterprises who value their humans and want to understand the future of Personal AI
- Yahoo, AOL
- Building tools for marketers
- AI = propensity marketing, used on large bodies of people to convince them to buy things
- He thinks of AI as "if you do this, you might want that"
- Ships a personal server, has your personal data, goes in your house
- Wants to have an AI that runs right on it
- How does it come together in products that people can use, understand, buy?
Robert Mithicki: cling.online
- Create a substitute of your best friend. Something personal to you. Always with you, always in the places where you are, runs on your mobile, helps you out, discovers the surroundings. One use case is the simple thing of the networking. If the best friend knows the people in the room he knows who to introduce you to.
- Have a privacy signal, akin to cargeigie mellon privacy beacon. Currently constrained by inputs, given AI they could make the signal more robust.
- Founded by Arlene Harris. Building family solutions, treating homes like small enterprises, they operate like a small enterprise, but without the technology. Automating the workflows in homes to help people live better lives and recognize the context of family. Built around personal goals. Users are in control of their data, their relationship.
- Created a device registry for kids. When you buy a mobile phone, when a phone connects to a network, the business can know that it's a kid.
Travis talks about architecture required to achieve an AI that could accomplish what we're talking about today.
Jobs to be done: what is needed to get it working? what do we want it to do?
- Sam: personal schemas. Wants to talk about data structure, not the things we could do with it
- Given what we already have, sensors coming online, etc
- Schema standardization first
- Understand things that stress me out before they do
- Would rather have data models that don't rely on schemas. Ontologies
- Smart toothbrushes or something?
- extend portability to extendability
- end game: all work together, not one central place for data
- Worthwhile to break apart the roles we wish AI. A lot of it is pattern recognition, another is deeper data analysis.
- End game: pick some low hanging fruit and eat it
- Johannes: first office manager he hired he had to travel, the next day she already figured out business insurance. She looked around, figured out what needed to be done, and did it
- That is what he wants
- Need to figure out how to collect enough information. Not sure if info already being collected is enough to understand what he needs for his daily life.
- AI will understand me in a ways that are predictive and not driven by him
- Need to find a way for ethical frameworks so analytic systems can recognize toward greater human well-being.
- Sari: works for IEEE ethical AI for human well-being. 3 year project, have done 1-year, have 300 different standards for human well-being. Doesn't even know how she would distill that.
- Wendell: Fiduciary crops up in these discussions. How a person or institution must act or behave. They don't generally extend to friendship, but do extend to well-being or in your benefit that you don't necessarily benefit, but need to do/know.
- Something that knows his ID's and when to renew them. Had to send a passport to a lawyer but it was expired. Someone who monitors spending. Has a kid on meds, doesn't trust anyone who is an MD, wants an AI who knows every human under my stewardship, flags for me if relationship between meds change. If it knows about him, can do stuff he doesn't have time for.
- Wants something to manage complexity. Settings/configure life to his personality would be useful. E.g., fix all the bad settings on his laptop. Docker for all his stuff
- Would very much like to see all analogies for who is wearing the collar. Where is the boundary of giveover where the AI puts collar on us and says "I know where to take you". Where is the governor or checkvalve there? *Who is in charge? How do you express guardianship, inertia?
- Would measure metrics of behavior. At start of relationship, more core human competencies you have, navigation. Over time, do you lose your ability to navigate?
- re. Chinese tendency to give over decisions: it's a culture thing
- Doc likes the metaphor of the collar, thinks agency is the way to frame it
- End game: we're all dead. Perspective. But in the mean time he wants to help out. Doesn't think there is an end game, there is a constant process that is a tug of war and we're al victims. Thinks it is cool that there are potentially competing companies here that all want to work together
- Any tool we build should be simple enough that a user should feel some sense of control over it.
- Doesn't want to have AI. Wants to become. Doesn't want to lose himself. Why can't he do that stuff on his own? Doesn't want to outsource it. When does the digital avatar become more you than you are?
- The right mental model is Douglas Englebart (inventor of the mouse): Humans and machines have to evolve together. As I learn more, my machine can learn more.
Frequent flyer miles / Loyalty programs
- Loyalty programs are a concrete thing you can do to help people. Many people love to collect stamps, miles, chips, the joy you hear from people who think they gamed the system, people love that stuff. Behind it is modeling to figure out when it's going to work and what the outcomes of the population are going to be.
- When we talk about AI, you can often just insert the name "God" and get basically the same outcome. First we had honeywell thermostats, now we have Nest. When you have to do something concrete, it often looks banal, like a thermostat.
- Sam: Martech papers are unbelievable. Just switch out who it is for and we can apply a lot of the science elsewhere for good.