1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
|
# bagel
well really it's a little hobby sentient-AI idea, but it's too small to do or be anything
# relevence
The idea is kind of that our minds have a very fancy network of nearness and farness of concepts.
When I think about something -- brainstorm on it -- my mind activates it and it activates relevent associated thoughts.
It pours through different thoughts that are nearby by association, and keeps the ones that are more useful, more relevent, while
discarding the ones that aren't.
# self-reference
Additionally, my mind when working on a task, can improve the actual process of working on the task. The steps can change.
The process for finding steps can change. The whole goal can change. I can even go out and wonder why I am working the task
in the first place, and think about that. Then I can act on the new decision.
By making code that writes itself, we move towards an intellect that can learn and improve.
# implementation ideas
I wrote out some quick relevence around how I might write code, and it seemed like the 'why' for each step was basically a small
chunk of code itself. It seems like there are relevent concepts in my mind that are roughly linked to these small behaviors
(habits or chunks of code). By making such links in actual code, the code looks more mindlike to me.
One of the keys is to get the code to make the links itself, and I think what is important there is the concept of patterns-that-
work. If you can discern when code works, then you can look for similar properties on the before and end states. If some of them
are always the same when it works, those are probably the ones that should be linked to the code chunk under study. This provides
for rudimentary learning.
# size
This code is written very tiny so that little work need be invested in it for it have conceptual growth. The idea and hope somehow
is that it move rapidly towards coding itself, such that as it gets coded, the developer need do less to work on it, being able to
instead make use of its internal functions for producing code-like behavior.
# language
By prototyping in a flexible language like Javascript, design work may be hastened a little. I also theorize that a good AI doesn't
need many cpu cycles to get stuff done, because it makes smart choices.
# etc
There is of course more. This writing content was made day-of-creation. Will I ever edit this again? Unknown.
|
