It’s starting to happen…
Call me psychosis or whatever, but am I the only one who views generative AI is more than the never-ending pop wars discourse on the internet, art, soul, consciousness, data, morality, yadayada but curiosity.
Every time a new model arrives, I’m not that excited it can do images better, movies, anime, or whatever. Instead, I think about how it’s getting us closer to understanding our world model, in the pixel realm or beyond the latent space.
Flashback: when I was still a kiddo in the 2010s, my dad introduced me to neural networks for predicting numbers in spreadsheets. Then came the stupid text bots. I still remember being in awe testing generative images in 2018 with waifu generator GANs during my nerdy chuunibyou highschooler era. In pandemic I still remember locked in home sign-in GPT3 for fun in the OpenAI Playground before ChatGPT born. Then the transformer and diffusion products dropped at scale in 2023, I knew the paradigm was going to shift.
Now, three years later: language & words have become really useful at doing executive tasks with Claude Code being mainstream, or OpenClaw, or Hermes, etc. Image models have been solved (currently the SoTA is GPT Image v2). Video models are almost solved (Seedance 2.0). World models like Google Genie or Odyssey already work well in pixel space. And fresh new paper, LeWorldModel with JEPA (Joint Embedding Predictive Architecture) in latent space is starting to actually work. So AI can finally predict the relativity of how physical law works.
I have this question: Will we arrive at an age where current SoTA models, LLM like Claude or Codex, video like Seedance, or world models, can run and generate in realtime directly inside your hardware, even on your phone, it’s become so cheap? Even in fiction, it’s hard to imagine the abundance scenario. Monitoring this advancement situation has been fascinating yet terrifying at the same time.
Your moat will be your metacognition btw, your ability to focus, your ability to taste, to synthesize & understand, cranking up your agency, building feedback loop systems to strengthen your cognition. And your currency will be your network outside your own neural network, good relations, a friends.
Welcome to golden age of doers.
Artists will still learn to draw, singers will still learn to sing, farmers will still learn to farm, engineers will still learn to build, brains will still learn to think, hearts will still learn to love.
The internet is no longer dead, we’ve been resurrected in its afterlife. Either hell or heaven, we’re heading in a beautiful direction, as the world intended, and in the end everything’s gonna be daijobu.