Every Take Is Now An AI Timeline
“Prompt Engineering” is just a way to talk about wishing things into existence, and we’ve been doing it for years.
OK - so you’ve heard of DALL-E or Stable Diffusion or GPT-3 - wait I mean ChatGPT - err, GPT-4, maybe LLAMA? or i̧̲͇͛̓͗͗̓͢ͅń̡͙͖̗̙͑͐̏̃͗̑͜͟ḑ̡̗̰̱̻̦̳̥̤̞̰͖͕̇͒͑͊̎͛̊̒̽̀͘͘͡͠e͓̯͖̲̎̒̎̇ci̧̠͕̯̝͊̀̈́͞p̧͕͉̰̭͔̞̠͌͌̌̑͘͞͞͡ḩ̻̦̖̣̣͈̖̟̦̟̟̹͉̩̠̟̞̝̱̾̑̓̒̓͗̓̂͊̽̏̽̏̾̄͌̒̍̔͞ȩ̨̢̧̙̱̥̤̣͎̜̹̫͓̼͎̯̠͍̤͇̑̏́̊̾̌͌̀͆͐̈̓͊̊̆͂͒̾͡͠͝r̨̧̳̖̩̦̻̰͔̜̫̀̎͛̑̂̎̋͂̑̾͝a͎̤͔͕͊̇̔͂̎ͅb̙̹̬̠̭̫͙̯̬̽̏͗͛̆́̆͂̇̐͘̚͜͜͜le͓͡ ? This is only the beginning.
Maybe you’ve seen the memes on Twitter about RLHF (extra schooling for badly behaving LLMS) and how GPT can be jail broken. Someone around you has engaged in Discourse about how AI will or won’t take any/some/all jobs. You received a TOS update email for a popular social media platform that you don’t read that allows them to vacuum up even more of your data to use for training their own models (probably not even that usefully or effectively but corporate FOMO is real).
Either way you have degree of awareness about TCOLT (The Current Or Last Thing)((AI)) and you have some opinion on it, one which is totally not related to how your feel about your profession and/or your generation. Here’s your thesis statement: pretty much any discussion that predicts the future is now implicitly a take on AI takeoff curves. We should talk about what happens after that though.
If you have seen the modeling/predicting the ai/rat community have done you can roughly break down these predictions into “When” and “Fast or Slow”. When is just “How many years do we have left before we hit the singularity” or various other thresholds. The Speed is the question of if one day we’ll wakeup to a swarm of nanobots dismantling the planet, borne from complex DNA strands emailed to some hapless lab or…if it’s “just” a smooth more human scale progression of people building better AI models over and over until Elon Musk releases a swarm of nanobots to dismantle the planet while beaming various roflcopter meme sounds and images straight to your dying brain.
Lots of people have various opinions about these timelines but it goes beyond the explicit statements about when we’ll celebrate Second Christmas. Aside from the worldlines where we fall below our current level of technology, the ability to ensconce Mind into Rock will inevitably change everything. Memetically, AI will take over long before it actually takes over (probably)((Even in the most boring dystopia outcomes)). Etc Etc, [[INSERT AI MATTERS ARGUMENTS HERE]].
A friend of mine in the Bay has been talking about how difficult it is to have conversations in Tech/Rat scene without people talking about TCOLT. People have started predicting it but the coming wave of AI strongly entering the Main Global Discourse is going to be overwhelming and so so Bad. lmao. Sorry pal.
There are many predictable responses people will have to AI discourse, all of them beaten to death around the AI/Rat circles, but I think the most notable one in a Memetic Sense is Degrowth. Degrowth is roughly, our current consumption is too high and we should go back to simpler, less consumer-capitalist driven lives. Various degrees of this span between “consume less pls” to “The Butlerian Jihad Didn’t Go Far Enough”.
Degrowth is DOA as an actual approach to handle Tech Progress - The Incentives are All Wrong And The Possible Outcomes Are Bad. The capacity needed to successfully reengineer our global economic system to provide for everyone is the same organizational capacity you would need to improve what we already have.
Although the graph doesn't accurately represent my issue with Degrowth, it can still be used as a fitting illustration for the issues I have and the sentiment we can recover from it. Every possible catastrophe is now part of the potential action space, and all have to be addressed. The analysis of the combination of AI, climate change, nuclear weapons, biological warfare, and more must be in terms of "how much" rather than "what if."
Degrowth proposes that we revert our civilization to a time before such catastrophic possibilities existed, but achieving this without a reality-breaking intervention seems challenging. Currently, our ability to create destructive tools surpasses our capacity to build, repair, and maintain them—unless AI technology proves beneficial. Implementing Degrowth would only postpone the problem for future generations who, due to the persistence of inherited knowledge, might face even tighter deadlines and struggle to adapt.
In my opinion, the widespread adoption of open-source models is a more favorable scenario, as it avoids the need to address the challenging issue of single-agent alignment. The more people who possess AI, the more competition and regulation will arise, keeping everyone in check. Don’t come for me.
Complaints aside, I think there is a good sentiment in the memeplex - that we should insist on things being a certain way about the world. This doesn’t require us to regress though, we can demand/prefer/request more particularly now. We live in increasingly cookie cutter apartments with the same products and appliances, decoration and customization in the Real discarded for customization in the Digital. When you don’t have to actually do the work, you can afford to ask for more. Basically, develop taste and an abundance mindset for transitioning into the most abundant possible world.
The coming discourse about this is set up to be the Eternal September Problem of a discourse to beat all others, magnified to literally super-human-population levels by the very thing we’re all going to be discussing. Current LLMs and other models alone, we will collectively process and reorganize around over the next few years. Politics will be predictable, news will be predictable, lots of orgs will be shaken up and die. People will be saved, people will get hurt. The internet has inured us to this, and there are a lot of people lined up to go through the cultural waves the hyper-connected have already become familiar with. The Doomers, the Deniers, the Copers, the Worshippers and the Criers. It will be important for you to remember that the things people post on the internet are Just Noise, and you will learn that lesson via computer generated means one way or another.
The actually interesting bit though is the question of what follows. The singularity arrives, so what?
Let’s say we survive and even get to Choose. The alternative being, we survive but the AI decides what is best for us. When we get Well-Organized Minds, who gets to drive the Balenciaga? There could be multiple levels to the singularity, where instead of a smooth progression up, the ascended minds hit a plateau that is difficult to breakthrough (Purple). We could of course, just have a Fast/Slow (Red/Orange) take off scenario and GPT is just our nascent godling, or maybe there is a Limit on useful intelligence that can be organized by a “single” mind. Maybe it’s Really Slow, and we just never get a superhuman hockey stick.
Either way, whatever we dream about here and now will drive the future that we end up building. “Prompt Engineering” is just a way to talk about wishing things into existence, and we’ve been doing it for years. SciFi is so predictive of future developments because our focus as a Culture is so driven by the small number of prolific writers who actually publish their ideas. The fun thing there is that the entry threshold for being prolific is dropping. Trust in widescale info-feed blasts is also dropping, but this just makes it local. Touch Grass, support your friends. Pitch them on the future, because it matters. What you choose to talk about, to generate, to push into the Feed, becomes the training data for people new to a space. Your bits of entropy nudge the distribution of new minds towards the reality you speak from within.
This discourse is gonna be a doozy and ideas and concepts are going to be warped beyond belief. When it happens, it’s going to happen fast and the internet will leak out into the physical world and change things in dramatic ways.
¯\_(ツ)_/¯
I’m an optimist.
-Hek