✉️ Dean's List #004

Hello my fellow analogians!
(assuming my human:bot ratio is good)
Today's issue is about... you guessed it... AI!... (cheer? yawn?)

Obligatory small talk:
Yesterday was Daylight Savings Time, the holiday where the morning steals an hour of sun from the night. This triggers a ritual where most people go around their house and change their clocks. I don't. I prefer to live in limbo for a few days. The dissonance keeps me on my toes. Either I'm happy to 'gain an hour back,' or I trick myself and show up to things early (I'm usually late). It's a win:win.

Meta-musing:
This newsletter shifts form again. Last week I sent a list of Google Docs and gave people a glimpse into my in-progress essays. People told me they loved the concept, but it would take a lot for me to commit to a form, especially when I'm only four editions in. This thing is like liquid. It's fun to run. It can and will keep mutating, until there's a good reason not to.

Today's main course, AI:
I published about AI twice last week. One was a compressed high-value Twitter thread. The other, a poetic ramble on a typewriter. I feel a long-form essay brewing on AI, and this edition is the first step to flesh out some ideas. Below's list is a 'triangulation.' I start with two links to published work, and then I riff on a bunch of related ideas. The end result is a public mosaic. Curious to see how conversations from this post can steer the direction of the essay.

1. My thread on how POP Writing lets you rise above the coming wave of infinite content – #TWITTER –  To summarize: Writing that is cheap, templated, and generic is at risk, but writing that is personal, observational, and playful (POP) is harder to replicate. a) AI can regurgitate facts, advice, and events, but not your own experiences. b) AI can generate rapid insights, but has no intuition to optimize around the emotion of surprise. c) AI-generated voices are either hollow averages or precise replicas, but your own experiments in language are unpredictable. Writers who love the craft have less to worry about.
   
2. Automatic intelligence – #TYPEWRITER – The core idea here is that AI is just the machine version of 'automatic writing.' This is more than just 'free writing.' At its most extreme, it's when the subconscious mind takes over, and your hands move automatically, forming language without you even thinking, as if you're in a possession state (legends exist). More realistically, it's about tapping into the subconscious. It's when you let associations spontaneously emerge and guide the form of the essay. I want to create a bridge between machine learning and our occult past of automatic writing from the subconscious.
   
3. The analog beauty of surprise – #IDEA – If comedians optimize for laughter, then writers optimize for surprise. Whether you're writing or reading, there's an emotional response to 'whoa moments.' Eyes widen. You pause. 'This could be something...' Machines can't have whoa moments. They can only use probability to determine what should come next. Sure, they can spawn millions of insights in seconds, but they don't have emotion as a barometer to measure intangible power. Intuition for the win.
   
4. The growing gap on the Internet – #IDEA – As AI advances, it's going to create a big gap in writing quality. It's going to consume the content-farm quantity obsessed writers, and it's going to make the good writers with soul even more prolific. It might dissolve a million jobs, and create a thousand Tim Urbans. It might spawn both a flood and a Renaissance. This growing gap is something we're seeing across the Internet in general. 100s of millions get sucked into manipulative algorithms, while a small group of digital pioneers unleash the sublime power of peer-to-peer networks. AI is going to amplify this divide even further. There is going to be a greater and greater pull to slip into a customized procedurally-generated hallucination where all your friends are bots, and a greater and greater reward for those who dive into the Internet with will-power and vision.
   
5. Autonomous entities as AI – #IDEA – It's weird to think of all of the 'entities' that spring from the subconscious as biological forms of AI. Dream characters, ghosts, DMT machine elves, imaginary friends, alien abductions, tulpas, Carl Jung's 'Philemon'. All of these 'spirits' are trained on your own life's experience (plus mythological archetypes), and then seem to take on a life of their own.
   
6. Recollect, AI as a Second Brain – #APP — Most of the AI apps we see are just prototypes. They basically take one feature (ie: the lasso tool), and present it as a full solution. But in the future, we'll see the rise of software suites where you have dozens of AI-powered tools. AI for titles. AI for rhythm. AI for word-choice. I'm super excited for an AI-based research tool. LEX is actually terrible at this. It presents me facts that sound real, but are comically fake. Recollect is way better for research, though the UI still makes it cumbersome. The concept is neat. It accesses your browsing history (sort of creepy), and then presents you excerpts with links as you write. It reduces the need to 'maintain' a Second Brain through linking and organizing. It even reduces the need to highlight. Fragments from your direct sources emerge as they're related to the present moment of prose creation.
   
7. AI as a live improv performance – #IDEA – There are parallels between AI and improv. Similar to GPT-3, Jazz is an unpredictable march forward, generating one bar at a time. There's both an awareness of a deep past, as well as a looping of patterns from the recent past. This concept is already real when it comes to AI-generated live streams for music. DADABOTS have 5 live-streams that have been jamming for months (some up to 3 years). Imagine if writing turned into a 24/7 live performance? We already have AIs that can re-create the voice of a famous author with near-perfect accuracy.  What if the collective intelligence of Twitter was plugged into a Hunter S. Thompson bot? It would be a riot. As trends and posts surface on social media, you'd watch Hunter S. Thompson live-write the history of the species.
   
8. Hallucinating words – #IDEA – Machine AI has a clear output... the screen. How is biological AI visualized? Can words be seen in the mind's eye? It's tricky. Language isn't baked into the mind since it's an external technology. It's way easier to see sights and hear sounds in the hypnagogic state before bed, since perception is our natural mode of input/output. But if you engage with symbols for long enough, it shapes the imagination. If you've seen Queen's Gambit, you probably remember the scene where Beth is lying in bed and hallucinating chess pieces moving on the ceiling. Obsession leads to a high volume of 'training data.' This happened to me once. Last November, I restructured ~20 long-form essays in one day (around 45,000 words). It was a huge mental load. For a whole day, my mind had a single function: Read words > rearrange words. That night, naturally, I hallucinated words. I saw sentences being typed out to me with my eyes closed. The thoughts of God? Unfortunately, they were gibberish. But what stood out to me was how lucid the words were. I was reading them as if I my mind was plugged into a Kindle.
   
9. Michael Dean as an app – #IDEA – After I edit someone's paper, they often joke that they wish I were an AI so I could help them 24/7. I'll admit, I think this is possible and not too far away. Things like structure, rhythm, word choice, and voice all have an underlying technique. They can be rationalized through parts and patterns. I've realized that the work I'm doing with writing curriculums and frameworks could serve as the foundation to build an 'AI editing app.' This cohort, I've built a 10-part framework to analyze an essay, and each part can be evaluated on a 1-5 scale. I'm currently gathering excerpts and refining the rubric with examples. The goal is to improve our editor training. And eventually, if we build an accurate model to diagnose student writing, we could pair them with exercises and content that meets them exactly where they are in their journey. It could lead to warp-speed, 12x, Benjamin-Bloom-grade learning. I'm not sure if I'll ever actually contribute to AI software, but if I do, I like the idea of it being an editor / assistant / tutor. These tools don't have to replace writers, they can educate writers better than anyone else, and help them express what's locked inside them.
   
10. Becoming immortal through AI – #IDEA – Art is a medium for artists to become immortal. Their creations outlive them. AI changes the way that artists survive. Instead of passing down their artifacts, AI helps them pass down their voice. I wrote about this a year ago, in Writing as Life Extension. I'd like to write 10 million words throughout my life, so that I can train and build an AI that closely resembles me. My great-grandkids might have a chatbot that encapsulates the voice, ideas, and essence of one of their ancestors.

Looking for:

• Feedback: Which ideas resonate? Which don't?
• Links: What should I add to my reading list?
• Ideas: Do you have original ideas here?

Thanks for your replies! I consider anyone who jumps in to be a collaborator on this essay, and I'll be sure to feature you in the published piece.

Until next time,
Michael Dean
michaeldean.site