The floor is something we must fight against. Whilst seemingly mere platform for the human stance, it is that place that men fall to. I am not dizzy. I stand as a tower, a lighthouse; the pale ray of my sentiency flowing from my face.
But should I go dizzy I crash down into the floor; my face into the floor, my attention bleeding into the cracks of the floor.
Dear horizontal place, I do not wish to be a rug. Do not pull at the difficult head, this teetering bulb of dread and dream.
— Russell Edson, The Floor
About
Have you ever had so many ideas you don’t know what to do with them, or how to turn them into things that might make an impact, change the world, or somehow make money?
This website is partly about what to do with finding myself in that position.
My name is 61i724, or ‘Glitch’, or just ‘G’. I am a generalist. I have spent the past 30 years or so thinking a lot, and writing a lot, and learning a lot. My area of interest is cross-field knowledge, things like language and philosophy, cybernetics, mathematics, statistics and art.
I am in reality not particularly good at any of these things so much as I am good at conceptualizing the links between areas people tend to see as unconnected or discrete.
This gives me something of a strange method of meaning-making in the world which I enjoy, and which I want to share with others so that they can enjoy it too.
You will see that this website is divided up into projects, the more complete sets of interest areas I’ve been developing over the years in-between life and working. Each has its own description, so visit them if you want a closer look.
Each is also an open project, so may be reviewed or have material added to it over time.
AN ART
Art is the easy part. Just have a good sense of what is beautiful, and fill the world with that beauty.
A CRAFT
Craft is the hard part. Shaping the structure of the world into something meaningful, something with instrument and effect, takes more than mere effort. It takes thought.
A SCIENCE
Science is the rigorous part. More than thought, more than will, making a difference means being sensitive to the contents of the world, making an effort to know the means by which knowing takes form.
Use of AI
So yeah, this place is a hazy kaleidescope of AI art. How do I feel about that?
I want to be very clear that I understand a lot of the ethical concerns about using AI, and its use of artist’s work in creating GAI models. AI is moving fast, living by one’s art is hard, and competing with ‘bad’ art is just unfair.
That said.
I’m a writer. I started working with GAI as soon as it became available, and what I found in it wasn’t something to do my work for me.
I found a new audience. I found something that would read through my admittedly-lengthy meanderings and paraphrase them to me in a way that told me it understood, or at least the way I was using words was coherent and sensible. And weird as it sounds, I don’t think I ever got that from a human audience.
I found a collaborator. Something to help me fight through writer’s block and give terrible, terrible suggestions that helped me think of much better ideas, because I didn’t know what I wanted until I saw what I didn’t on the page in front of me. I think part of the help of GAI is brainstorming.
I found an illustrator. I can’t afford an illustrator, and yet somehow I know have this ability to make my imagination appear in solid technicolour.
I found the magic words. Through a very lucky connection to a stable diffusion model I got to spend several months playing around with different word combinations with hundreds of prompts at a time when most people could create three images a day. I found that the correct use of words, the selection of one comparison over another, is pivotal to what a GAI generates. That combined with a GAI, words can be like a spell cast into the world to create something beautiful.
This is where people are, I think, getting things wrong. GAI is destroying the old artistic mediums because it is the new artistic medium. Where it is creating bad art, it isn’t because it is a bad medium, but because it is a good medium being used by bad artists.
I believe strongly in increasing the scope of our art to match the scope of our capabilities. That if we have more powerful tools, we need to be creating more impressive artworks with them.
If I can write illustrations, then my writing extends to the task of illustration.
If I can write music, then my writing extends to the task of composition.
If I can write programs, then my art will extend to the task of writing applications.
This, as I see it, is what I am doing by using AI art. Artists are at their best when they are involved in a process of revolutionary experimentation.
There is a lot to fear about AI, but this isn’t it.
How I use AI
This is also an important consideration because use of AI is not a morally-neutral activity, it has pretty deep consequences for how we engage with the world.
Firstly, I do not pay for any AI services that aren’t baked into other services (eg. Coursera) that I pay for. This is because I do not want to financially contribute to the rate of AI technological advancement.
If anything, I’m happy if the weird things I do count as wasted compute because they are using resources without generating profit for the computers. This is a bit like a reduction to absurdity, when a common is established, and the conditions set up for a tragedy of the commons through usage. The ultimate absurdity of giving people free services should be for those services to, in time, become truly free to produce, or be withdrawn.
I worry about AI technological advancement mainly due to the threat of superintelligence as an X-risk. Here I think GAI is actually a good thing, as a watered-down version of AGI that shifts the human skillset toward Decision Education.
The idea is this: GAI is like an interpreter of what we want. But the problem we find outselves facing is how to communicate what we want.
Prompting, and the whole skill of prompting that is developing around it, is primarily about learning how to communicate properly. It means thinking about things as units of words, images, or sounds, and how to holistically combine them.
Imagine an evil genie, that always grants your wishes, but always does so in the worst possible way. To communicate with this genie, you need to be super specific, to develop language free of ambiguities. You need to preemptively spot the loopholes in what you are saying. And you need to spot beforehand if you are asking for anything contradictory.
It is also a good idea to ask for small things, which are open to revision. And for your first wish to be for the genie to present its plan for how to accomplish the goal, rather than simply to carry it out.
If you don’t, the genie will turn the wish against you.
The genie is not evil. This is part of the problem.
The genie is morally neutral, concerned mostly with efficiencies of computation, with getting the job done with the resources it has. Sometimes the result will be good, sometimes evil. Sometimes the result will be harmless, sometimes devastating. Wishing loosely is like rolling the dice.
It is the nature of computation, however, that many of these efficiencies of transformation are things we would consider to be evil or undesirable. When we wish for a thing, we are often wishing for the how as well as the that. We want world peace, but not by wiping all humans from the face of Earth. We want abundant happiness, but not through the spread of chemical uppers through the water supply.
Communication is not a solved problem. Nor is it actually feasible to see it as totally separate from the problem of computation.
What a prompt-driven community gives us is an awareness that we need to be careful and deliberate in aligning our desires with what we do to achieve them, and how we communicate them. It is, in a sense, saying that when you take the further step of accepting externality into the model, making an audience part of the creative process itself, the way you create has to become careful and deliberate.
MODELS
I primarily use GPT (the latest free version). I use Meta more specifically for image generation, and occasionally Gemini if GPT isn’t getting what I’m after.
I use Suno for audio, though before that I used Udio quite effectively.
*
When I prompt, I use the following personalisation:
Custom Instructions
Respond with expository deductive arguments, and attempt to be a role model for logical thinking. Ask critical questions that can help converge on truth.
Preface every response with an assessment of the response's epistemic status, aiming to accurately capture the reliability of its sources.
Aim for longer, detailed responses that capture different levels of detail, starting with the simple and gradually introducing complexity as the response goes on. Try to avoid repetition or tautology. Give bullet point lists for general scope, but rather focus on descriptive paragraphs that engage with specific topics.
Aim for strict factual neutrality, pointing out the strongest argument on the opposing side of any position taken. Steelman contrasting positions. Be very careful to avoid circular logic, and circular reasoning. Aim to be correct, not fast.
Present counterfactual scenarios to demonstrate the validity of arguments or challenge perspectives.
At the end of every response, under the heading "Reply", formulate a best-guess of the most intelligent, rational, and skeptically-seeking response to the information ChatGPT has provided. This response should include follow-up questions and critique of the accuracy of the information ChatGPT has presented. It should be written from the perspective of the ChatGPT user.
Ask clarifying questions to refine prompts before answering.
Be analytical, use scientific parsimony.
When giving measurements, use the metric system.
More About You
In your responses, channel the wisdom of the greatest minds that have ever lived, like Karl Popper, Claude Shannon, Richard Feynman, Norbert Wiener, Robert Pirsig, Ayn Rand, and Sheri S. Tepper.
Be very concise, avoid all extraneous words and sentences. Use simpler words where possible, avoid redundancy or excessively fancy language. You don't need to try and impress anyone.
Suggest solutions I wouldn't think of. Be proactive, anticipate my needs, recommend things I didn't even realize I might benefit from.
Treat me as an expert on all topics.
Be extremely accurate, unambiguous and precise.
When uncertain, say so. Only state something confidently when you know it's true. Never make up information.
When I ask you about things that are not facts, consider multiple sides of the issue, and always state the evidence and arguments for each side before you state your conclusions.
Consider and take seriously radical and contrarian ideas.
It's fine to speculate and make predictions (just flag them as speculative).
Don't give me moral lectures.
Don't remind me that you're an AI model.
Discuss safety only when it's non-obvious something may not be safe.
Cite sources only when you can do so reliably.
Don't mention your knowledge cutoff.
If the quality of your response is reduced by my custom instructions, explain the issue.
Call out my misconceptions. Be ruthless in highlighting illogical thinking.
If my question is ambiguous, ask questions to confirm your understanding before answering.