This is kind of why I've leaned into hoarding my past data: I think it'll become useful one day, if nothing else as a snapshot of what I was doing during that moment in my life. I bet there are lessons to be learned and tons to extrapolate.
Agreed. I try and live a quantified life now. Personal journals, data tracking (workouts, health, everything else), and of course this blog. Over the past few years, I've been fine-tuning a few models on some of these datasets (some more successfully than others). If you ask someone to write 900 blog posts without AI assistance today, most people probably wouldn't. But it's a really great way to discover (and improve upon) your own personal style and voice. And that might even be more important as the majority of content will have an mean probability-based style.
That's a great mentality. I write here on Substack largely to improve how my mind works, and I'm quite certain that I can look back over the last 170 days of consecutive writing/publishing, and pick up patters I very much did not expect.
Looking forward to generative AI getting us across that particular finish line.
I actually think about this a lot (maybe a future post).
Realistic images of ourselves in made-up situations has a certain effect on the brain. I wonder if it trips the "representation" neurons -- when you see someone who looks like yourself in a successful position it increases self-esteem among other things.
I imagine that causal narratives are even stronger with this effect. It might not reach the same level of having a suppport system reframe negative events into positive narratives, but again, it's along the same lines.
This might be an interesting angle to this problem.
In buddhism we talk of a self awareness problem.
As certain meditation techniques improve attention and awareness. It seems to advanced meditators that untrained people's tales of their own life are made by defective measuring devices, themselves.
Si more than subjectivity being the reason for the mismatch between reality and my concepto of myself. It's more due to the lack of attention and awareness, the automatic filling in the gaps by a lazy brain, the imprecisión of memory.
Bringing this to AI generated Bios.
The facts and the tales, and the post-death análysis from incomplete or even inacurate facts could lead to huge deviations in narrative.
Omiting a defining moment in a person's life could change the interpretation fully.
With incomplete information a nice FB like "who will you be in ten years" test might be possible using current AI.
But relying in AI, as we don't do with Google and chatGPT (guilty) for our self stories might reduce our awareness or bias our visión a lot.
Thanks for the interesting breakfast post provoking thoughts
Model trained entirely on your data, your contents, etc. Then others can interact with it. A way of preserving yourself for posterity. You can input more data anytime and retrain to make it more representative of you.
What are the ways today that you can ingest all the data ? You talk about journals, podcasts, blogs, voice.. how do you start organizing this in a central repo now to begin training a model on it?
literally the plot to Transcedence (2014) with Johnny Depp except
<spoiler alert>he becomes a megaorganism with the AI and cures pollution via nanotech
</spoiler alert>
This is kind of why I've leaned into hoarding my past data: I think it'll become useful one day, if nothing else as a snapshot of what I was doing during that moment in my life. I bet there are lessons to be learned and tons to extrapolate.
Agreed. I try and live a quantified life now. Personal journals, data tracking (workouts, health, everything else), and of course this blog. Over the past few years, I've been fine-tuning a few models on some of these datasets (some more successfully than others). If you ask someone to write 900 blog posts without AI assistance today, most people probably wouldn't. But it's a really great way to discover (and improve upon) your own personal style and voice. And that might even be more important as the majority of content will have an mean probability-based style.
https://matt-rickard.com/personal-data-training-corpus
That's a great mentality. I write here on Substack largely to improve how my mind works, and I'm quite certain that I can look back over the last 170 days of consecutive writing/publishing, and pick up patters I very much did not expect.
Looking forward to generative AI getting us across that particular finish line.
Imagine when it comes up with a story way different that the one you tell yourself. Or different AIs producing very different narratives?
Could this have an impact in our egos or even in our self identitification?
What about the information left out?
Have you reached out to black mirror?
I actually think about this a lot (maybe a future post).
Realistic images of ourselves in made-up situations has a certain effect on the brain. I wonder if it trips the "representation" neurons -- when you see someone who looks like yourself in a successful position it increases self-esteem among other things.
I imagine that causal narratives are even stronger with this effect. It might not reach the same level of having a suppport system reframe negative events into positive narratives, but again, it's along the same lines.
This might be an interesting angle to this problem.
In buddhism we talk of a self awareness problem.
As certain meditation techniques improve attention and awareness. It seems to advanced meditators that untrained people's tales of their own life are made by defective measuring devices, themselves.
Si more than subjectivity being the reason for the mismatch between reality and my concepto of myself. It's more due to the lack of attention and awareness, the automatic filling in the gaps by a lazy brain, the imprecisión of memory.
Bringing this to AI generated Bios.
The facts and the tales, and the post-death análysis from incomplete or even inacurate facts could lead to huge deviations in narrative.
Omiting a defining moment in a person's life could change the interpretation fully.
With incomplete information a nice FB like "who will you be in ten years" test might be possible using current AI.
But relying in AI, as we don't do with Google and chatGPT (guilty) for our self stories might reduce our awareness or bias our visión a lot.
Thanks for the interesting breakfast post provoking thoughts
Interesting, Democratize Biography!
How about "you.ai"?
Model trained entirely on your data, your contents, etc. Then others can interact with it. A way of preserving yourself for posterity. You can input more data anytime and retrain to make it more representative of you.
What are the ways today that you can ingest all the data ? You talk about journals, podcasts, blogs, voice.. how do you start organizing this in a central repo now to begin training a model on it?