I've been thinking a bit about solving this same problem, but I feel there is too much overhead to log things manually.
What we need is better interfacing with computers so that we can just think "log my weight is 190 lbs" or "log I ate 1 small chicken breast and a cup of brocolli". Voice control and my phone are not sufficient for my needs so far. I also want to be able to have full control over my own logs in plaintext.
I built a natural language health and nutrition journal for Telegram and SMS, but once it was working I didn't want to use it. I'm not convinced that keeping a meticulous log solves a real problem. Loosely structured journaling is definitely good, but not because of the data it encodes; it's benefit is the self-awareness that arises from putting thoughts and observations down in writing.
I think software people over-estimate the usefulness of software for guiding personal decisions and behaviors. There are far better ways to gain insight and control over your impulses and habits, ex. meditation, conversations with a friend, a skilled coach, community involvement, reading books... The one exception might be medical. Biometric logging could improve medical insights, but that carries a big vulnerability to exploits on that data.
Safe, simple, cheap, non-invasive neural interfaces with a good software ecosystem might be the next big wave of human advancement, with the wide adoption of the internet as the previous one. It opens the door to so many possibilities and use cases.
> Every important system has logs since they’re necessary to improve the system over time. That brings me to the most important system I know. One that we never designed - but use all the time. The human body.
I would argue that there are two more important systems:
- Planet Earth
- Human societies (and overall humanity itself)
Writing a similar essay but with these as the the subject would be a difficult but fun and potentially fruitful undertaking.
I wonder: how might one go about this, considering the complexity of the system?
These folks seem to me to be the least, uh, starry-eyed of the would-be Foundationites, but my knowledge is not comprehensive, eh? I only know about these folks because I'm interested in their underlying graph-based Prolog-implemented database, TerminusDB. ( https://terminusdb.com/ )
- - - -
FWIW, I view the Internet et. al. as a kind of GAI with humans for neurons. From that POV Seshat is one of many ways in which we/it are becoming self-aware.
I've been thinking about this on and off and at the moment I'm feeling like automatically collecting log information is a good approach, at least for me, so that I can set and forget until I'm ready to explore. I really like the way that smart watches are making this easier.
I use an iPhone and I've been wondering whether there are any good tools that can automatically collect some information about my activities on the device, sort of like the screen time utility-- off the top of my head:
* send an email
* time notetaking
* time and address of web pages visited...
Keeping a uniform activity log of what I'm doing on my devices would be something that I could use to go back over and figure out what I've actually accomplished, review it against my goals...
Thanks for sharing! This resonates strongly with me.
One thing I did (and fairly recently too: https://github.com/alefore/edge/commit/6e93c2bb327f57d13d067... says just 10 days ago) is adjust my text editor (http://github.com/alefore/edge) to keep a log per file of all the operations I do on that file. I expect to use this to generate reports along the lines of "what files did I work on last Thursday", "how much active (for some definition) time have I spent modifying this file" or "on how many different days have I opened this file". Perhaps I'll also use this as a cluster to feed to the Bayesian filter ("I tend to edit file X whenever I edit file Y") that I use to optimize my autocomplete predictions/search history.
(I also plan to use this to see if I can find ways to optimize the types of transformations that I bind to my keys if I notice special patterns (e.g., when I do transformation X, I very frequently do transformation Y immediately after), but I haven't yet thought very much about how I'll do this.)
I'm somewhat curious as to how much value I'll actually derive from this. I tried the Quantified Self in the past but ... I think it suffers from the fact that reality tends to be too complicated to be predicted by ~simple models.
In my opinion this is the entire selling point/core value of Augmented Reality and is why it's been a pursuit of mine since 2006 or so.
And for clarity, it's the world facing camera that is the most critical part of AR, in my opinion, not what is displayed to the user. The display simply gives me most compelling reason to the user to adopt the technology. The reason for this is because a world facing camera system is a tractable tangible thing that we can build and actually deploy today. The current state of object detection with CV, and the other long tail analysis you can do with it, allows us to get really really good at making inferences of world and user state, based on detected objects (+ some telemetry).
So today, you can get really close to a high resolution log, from strapping a camera on your head and building an object detection pipeline. The challenge here though is of course what you all expect, privacy, monetization, influence, control etc... The world where everyone is logging everything about themselves, which means they would also be logging their interactions with others, is the no-privacy world that everyone is terrified of.
You can't create Laplace's Demon and also preserve the virtues of privacy.
If only it worked like a log, though. Unless you're a John von Neumann level outlier, you can't query your memory for "what's everything I did between 9 and 10 AM on October 12, 2017?"
Nothing. Just saying memory alone isn't sufficient for most people. Most would need to rely on something else in addition, like remembering to manually make note of things in the moment.
First, let me say that I'm not against e.g. "quantified self" et. al. at all. It's fine. (IMO) And writing has, in general, been a benefit, eh?
> Just saying memory alone isn't sufficient for most people.
I'm merely pointing out that the reason we (as evolving beings) have memory in the first place is due to the adaptive advantage of "logging" experience. ;-)
FWIW, memory has been sufficient for most people: we're here, therefore every one of our ancestors did at least well enough to make a baby (who grew up to do at least well enough to make another baby, and so on right up until you, today.)
Right, I mean memory is sufficient for most sorts of things, but if you want to do the sort of retrospective debugging and analysis the article talks about, you either need to be born a savant or rely on some external system. Eventually, when we have good neural interfaces, I think this will be done automatically, but that may or may not happen in any of our lifetimes.
Natural selection is good for ensuring you make it to a bit past reproductive age, but for everything else it's really hit-or-miss. As one would expect. We're just the first known lifeforms that seem to have an imperative beyond replication.
26 comments
[ 4.2 ms ] story [ 29.6 ms ] threadWhat we need is better interfacing with computers so that we can just think "log my weight is 190 lbs" or "log I ate 1 small chicken breast and a cup of brocolli". Voice control and my phone are not sufficient for my needs so far. I also want to be able to have full control over my own logs in plaintext.
I want to be a cyborg already...
https://github.com/lordi/tickmate
I think software people over-estimate the usefulness of software for guiding personal decisions and behaviors. There are far better ways to gain insight and control over your impulses and habits, ex. meditation, conversations with a friend, a skilled coach, community involvement, reading books... The one exception might be medical. Biometric logging could improve medical insights, but that carries a big vulnerability to exploits on that data.
https://quantifiedself.com/get-started/
I would argue that there are two more important systems:
- Planet Earth
- Human societies (and overall humanity itself)
Writing a similar essay but with these as the the subject would be a difficult but fun and potentially fruitful undertaking.
I wonder: how might one go about this, considering the complexity of the system?
The Seshat Global History Databank people are into this sort of thing (real-world psychohistory): http://seshatdatabank.info/
( https://en.wikipedia.org/wiki/Psychohistory_%28fictional%29 )
Now this looks very interesting, I am going to give it a read.
If you have anything else along these lines up your sleeve, I would love to see it!
These folks seem to me to be the least, uh, starry-eyed of the would-be Foundationites, but my knowledge is not comprehensive, eh? I only know about these folks because I'm interested in their underlying graph-based Prolog-implemented database, TerminusDB. ( https://terminusdb.com/ )
- - - -
FWIW, I view the Internet et. al. as a kind of GAI with humans for neurons. From that POV Seshat is one of many ways in which we/it are becoming self-aware.
I use an iPhone and I've been wondering whether there are any good tools that can automatically collect some information about my activities on the device, sort of like the screen time utility-- off the top of my head: * send an email * time notetaking * time and address of web pages visited...
Keeping a uniform activity log of what I'm doing on my devices would be something that I could use to go back over and figure out what I've actually accomplished, review it against my goals...
One thing I did (and fairly recently too: https://github.com/alefore/edge/commit/6e93c2bb327f57d13d067... says just 10 days ago) is adjust my text editor (http://github.com/alefore/edge) to keep a log per file of all the operations I do on that file. I expect to use this to generate reports along the lines of "what files did I work on last Thursday", "how much active (for some definition) time have I spent modifying this file" or "on how many different days have I opened this file". Perhaps I'll also use this as a cluster to feed to the Bayesian filter ("I tend to edit file X whenever I edit file Y") that I use to optimize my autocomplete predictions/search history.
(I also plan to use this to see if I can find ways to optimize the types of transformations that I bind to my keys if I notice special patterns (e.g., when I do transformation X, I very frequently do transformation Y immediately after), but I haven't yet thought very much about how I'll do this.)
I'm somewhat curious as to how much value I'll actually derive from this. I tried the Quantified Self in the past but ... I think it suffers from the fact that reality tends to be too complicated to be predicted by ~simple models.
And for clarity, it's the world facing camera that is the most critical part of AR, in my opinion, not what is displayed to the user. The display simply gives me most compelling reason to the user to adopt the technology. The reason for this is because a world facing camera system is a tractable tangible thing that we can build and actually deploy today. The current state of object detection with CV, and the other long tail analysis you can do with it, allows us to get really really good at making inferences of world and user state, based on detected objects (+ some telemetry).
So today, you can get really close to a high resolution log, from strapping a camera on your head and building an object detection pipeline. The challenge here though is of course what you all expect, privacy, monetization, influence, control etc... The world where everyone is logging everything about themselves, which means they would also be logging their interactions with others, is the no-privacy world that everyone is terrified of.
You can't create Laplace's Demon and also preserve the virtues of privacy.
Aw, c'mon! It's called memory and our systems spend a significant amount of ATP on it.
> Just saying memory alone isn't sufficient for most people.
I'm merely pointing out that the reason we (as evolving beings) have memory in the first place is due to the adaptive advantage of "logging" experience. ;-)
FWIW, memory has been sufficient for most people: we're here, therefore every one of our ancestors did at least well enough to make a baby (who grew up to do at least well enough to make another baby, and so on right up until you, today.)
Natural selection is good for ensuring you make it to a bit past reproductive age, but for everything else it's really hit-or-miss. As one would expect. We're just the first known lifeforms that seem to have an imperative beyond replication.