Why Birds of A Feather? (WIP)

Rock, Paper, Shears.
No, thats not right…
Rock, Post-It, Scissors.
Hmm, that's not it either…

But what if it could be?

The Burden of the Creative Expert

Creating is a process of conjecture and testing.

Conjecture is expansion -- the words on paper, the melody played on the guitar, the creation of "something" from "nothing".

Testing is contraction -- the edit, the reality-checking, the refinement of the raw material.

Conjecture and testing require different skills. One requires an open mind, a willingness to fail, a penchant for play in unknown (potentially dangerous) areas. The other requires a deft hand, the ability to cut and remove without damaging the core product.

All creators start out with great taste, but bad execution.^[1] Over time, through study and trial and error, skill is slowly gained. Habits and personal biases are formed, and these serve as the bedrock of the creative's ability to perform.

Over time, the creator gets better at conjecture: out of all the available paths, they choose paths that are likely to work, their ability to pattern match increases greatly.^[2] The amount of "divine inspiration" increases drastically.

Over time, the creator gets better at testing: their edits become more precise, the cuts leave less obvious signs, the "quality" of work increases drastically.

The burden is getting to this level of mastery. The burden is maintaining this level of mastery once it is achieved.

Inside The Mind There Be Viruses, Fusion, Staking

An idea is a virus. A novel (external) idea enters the pitch black skull through the eyes, ears, nose, or mouth and worms its way through the neural circuitry.

From there, the idea conducts battle. It is fiercely attacked by pre-exisitng biases, truthy filters,…

{todo}

If accepted, the idea fuses with other ideas, making its way into the fabric that will serve as a future judge of new idea viruses.

Are Embeddings a Point in Space?

{todo}

Same But Different

{todo}

What Everyone is Missing About Embeddings

As of January 2023, every usage of embeddings I've come across have leveraged a single facet of embeddings: returning the top result (or perhaps up to the top three).

This is a mistake. All embeddings, similar or dissimilar, provide unique information. By tossing out 99% of a dataset, volume is sacrificed for surface area.

{todo}

The 99% of the dataset is not just random information, it is weighted random information. This distinction might feel dumb, but it's critical to understanding the evolution of recommendation systems powered by AI.

All data put into embeddings represent a subset of the entire latent space of a model. These distances reveal themselves "deterministically". After data is clustered, a visual representation of idea space can be drawn visually, showing the self-similarity of certain ideas or trend lines.

{todo}

Chaos Is A Good Thing, Actually

Chaos in the space of ideas is conjecture. It's the ability to "jump" out of a local maxima in favor of venturing into the dark to look for a global maxima. Unfortunately, as a creative project begins to anneal, the idea space shrinks. The number of doors begin to close off to the creator, and the obvious choices become painfully obvious.

However, embeddings do not suffer the same weakness. Since all comparisons are done at the snapshot level, there will always be a lowest-match object. This lowest match object is the furthest thing from the pattern generated. This serves as a window. An out of a pattern that is beginning to self complete.

…unexplored space of co creativity with ai serving as an antagonist in a way that promotes better work

{todo}

Quadrant II Thinking

{todo}