Topics: Computing
The Problem is Revealing Order in Data
- a
- have you ever done a simple practice where you take a piece of content
- assign a binary to each word
- take two pieces of content
- compare then at binary level
- then evaluate how the results look like to the eye
- observe clinically so to speak
- if not
- we should do this
- the question is this
- patterns
- how can we by our own creativity and through manipulation of the data
- reach a condition where we can maximise the appearance of patterns
- it seems obvious that a system with only two states
- 0
- 1
- is the one with the highest number of identifiable patterns
- basically
- to say that
- that any data
- any format
- any size
- any data at all
- when represented in binary format
- (then the side question is how the conversion from other system to binary affects order of the output i.e. the method for conversion should be such that maximises order…which brings us back to the creative input we have to put in to this)
- so yes
- when any data is represented in binary format
- it will express the highest level of order
- compared to any other system of representing it
- that is the hypothesis
- and yes of course
- it seems obvious
- if that is the case
- the the question really is
- “which type of conversion from non-binary to binary results in the highest level of order”
- for example text
- there is virtually infinite ways of assigning binary IDs to words, phrases, etc. etc. etc.
- if it is correct
- that binary will always express most order
- which again, seems obvious
- then half of the problem is solved
- by focusing our attention on the other half of the problem
- we eventually arrive at the solution
- of the compression problem
- i.e.
- the optimal compression level for binary systems
- I think this might be the most interesting of all of the problems we’ve discussed
- it has even more profound impact than generic AI
- far more
- it solves a problem that then leads to increased ability to solve the generic AI
- generic AI is a subset of the compression problem
- what I like about it
- is tangible
- you have different types of data
- you compress them
- you execute them
- you compare them
- and that’s it
- as long as you went lower
- you were fine
- this is the interesting thing here
- the problem is not to make the data go smaller
- the problem is reveal order in the data
- which is great
- as we already identified
- that we have relatively speaking infinite number of different options to how we can do the conversion
- this is something the machine will be really great at
- basically we can have that as a goal (improved compression)
- and have the machine endlessly try different variations that lead to ever increasing levels of order
- as long as we have that, then we are fine
- so we do our part
- and the machine does its part