03 sep 2026
02 Aug 2026
Human-made intelligence does not exist. Certain corporations claim the opposite, to make their large language model (LLM) systems seem more useful than they are. Use of language is not evidence of intelligence; even machines can do it. Intelligence arises from an organism living in the world, it is not contained within disembodied language. There is no evidence whatsoever for separation of mind and body.
Most references to "AI" refer to internet LLM-based chatbots run by publicly traded corporations like Google, Anthropic, OpenAI, Palantir, etc. The corporations involved are directly and intentionally using these systems to dismantle civil, legal, and political systems in this country and others for their own gain, enabled by overt racists and fascists in political office. Those tools have become surveillance for the state and corporate profit. I will not use them in my work.
Writing software, and computation itself, is a human discipline, an art. I enjoy writing code. The entire microcomputer industry of the 1970s and 1980s was started by amateurs, often doing foolish things against the advice of those who "knew better". They did so by sharing and communicating work, in small businesses, hobbyist magazines, in clubs, at home and with friends, later online in bulletin boards, a nearly entirely amateur effort.
Work you do with an LLM is fundamentally not yours. You did not do the work, did not gain the knowledge, did not explore, discover, reject disused paths (this alone is much of "knowledge"!), did not discuss the ideas and methods in that code with others, did not involve the lives of others around you, but instead enrichened an exploitive corporation. The resulting work is derived from someone else's prior exploitation, made invisible. You do not have control of your work when you are reliant on the opaque, proprietary, and increasingly expensive use of a corporation's machines.
Everything that follows from the use of online chatbots is fruit of a poisonous tree, and any work that follows from it is equally irredeemable.
I will not knowingly use tainted program, code, or library my own work. If such is later found I will replace it. The issue is one of ensuring people can understand and maintain and learn and importantly, benefit from shared code. "Deliverable" is a business problem not a human one, and "faster" is a poor excuse.
Complexity in and of itself is not a thing to celebrate. Understanding, clarity and intent, being succinct, and community are things to devote energy to and celebrate. The world is built with human connection, not corporate tools of exploitation.
Last, I distinguish between large language model and similar technologies from the corporations currently wielding them in public. It may be that open and tolerably ethical training data exists, or could exist; that the LLM technology is fascinating and groundbreaking work. It is very likely these technologies are very good for at least some forms of code-writing qnd who knows whatever else linguistic-centered collections of data might benefit from deep and exhaustive interpolation. But interpolation of a language-based collection of data is "all" they can do. At this point in history we know that organisms, a living body in the the world, can extrapolate from knowledge and experience. It seems doubtful this will happen with any current technology, and impossible to happen in a stored-program machine with "lots of data".