The much-needed data sources to construct a personal AI assistant
How can AI leverage existing online data to understand us better?
Hello, it's Sheldon again with some quick thoughts. I'm writing on Substack to find my next job. Feel free to reach me on LinkedIn.
When the concept of a personal AI assistant was first introduced, my initial excitement quickly turned to disappointment. Why? Because AI didn't understand me well enough.
For an AI to serve as my personal assistant, there’re 3 key kinds of data AI should leverage on. All these data are available online and comprehensive enough to understand me fully as a complex human.
1. Content I’ve consumed online
Specifically, my browser history over several years, and all the posts I've consumed and interacted with on social media.
I'm fortunate with my browser history, as I mostly use Chrome and Google has made the browser history somewhat accessible to third-party apps. I'm pleased that the Heyday app is already doing this.
However, when it comes to data on Facebook & Instagram, we're not as fortunate.
At this point, I wish I could simply pay Meta to retrieve all my data from over the years. This includes all the posts I've consumed and interacted with, and the digital version of myself created by big data, the data created and used by algorithms to determine my next post and the ads I'll see.
By owning all this data, I could provide it to my AI assistant for analysis, helping it to better understand me better and act as my personal assistant.
2. My emails, meetings, and private chats on WhatsApp & Slack
Imagine an AI that could analyze all past messages I've sent to my ex-girlfriends and provide practical relationship advice for the future. Or one that could review my email history - newsletters I've subscribed to, emails I've opened versus those I've ignored, and meetings I've scheduled over the years.
3. Things we've purchased online
Nothing says more about us than the food deliveries we've ordered, the show and performance tickets we've purchased, and the items we've bought online.