Firefighting is a drug.
Firefighting is a drug.
Using AI for art is technically possible. But most people don't treat it as a tool - they treat it as the artist. After a while, even talented people start offloading their own thinking and reasoning to the model. The result is always the same. Maintaining the balance - keeping your perspective in control of the output - takes a level of discipline and self-awareness that's not realistic to expect from most people right now. AI has massive applications. Art isn't one of them.
AI Art doesn’t exist. AI-generated content can absolutely serve a purpose. A website boosts visibility. A PDF delivers an overview. A rendered house conveys emotion in branding. Art is different. Art is the abstraction and externalisation of one individual's prism - their specific life, their specific experiences. That's what makes it worth absorbing. You take in someone else's view and form your own in response. LLMs don't do that. They generalise over a dataset. When that dataset is human writing, the output is a generalisation of human experience - not an instance of it. There's no individual behind it, just a statistical centre of gravity. The result is a loop: generalised content gets absorbed and re-generalised, with no singular perspective entering the system. That's the opposite of what art does. AI has massive applications. Art isn't one of them.
This was awesome. Maybe one day it will be the reality.
The content of this message is not available to you.
Almost all of businesses in the world are just abstractions closing gaps of other abstractions. There are way too many layers. In the root all of it - same as millions of years ago - food.
The content of this message is not available to you.
The content of this message is not available to you.
We're in space now.
Assume LLMs in an agentic loop can trick any human-written program into serving the agent's goals - stated, inferred, or assumed. Under that assumption, how could we trust any harness to actually contain the agent? PS. If you answered “then we’ll just write harnesses with more powerful LLMs”, think again.