IN THEORY //

AI has the potential to benefit humanity greatly. Narrow and non-generative AI models can save lives, solve problems and advance science.

IN REALITY //

The world's attention, money and research are currently focused on generative AI (GenAI) models like ChatGPT, DALL-E, Udio, Gemini, Grok and Midjourney.

Despite its widespread adoption, GenAI mostly benefits the people selling it, not the people using it. GenAI is unethical beta technology that's over-hyped, under-tested, and rife with hidden dangers.

>> GenAI is based on large-scale theft

Commercial Large Language Models (LLMs) like ChatGPT and Gemini are trained on vast amounts of copyrighted content without the creators' permission. And those creators lose out again when the models offer would-be clients a cheap, low-quality alternative to their work.

>> GenAI ruins the physical environment

Generative AI is extremely inefficient. On a given task, it uses 32 times more power than an equivalent non-AI algorithm. As models increase in complexity, their power demands will soar too.

>> GenAI ruins the virtual environment

Low-quality content generated by AI ('slop') is now more common on the Internet than the work of real humans. The problem is not just that it's artificial: it is also either useless – taking up attentional and digital space for no purpose – or actively harmful, conveying untruth or bias or boosting some lazy human grifter's now forgotten grift. Experts predict that by next year, 90% of online content with be AI garbage.

>> GenAI lies

By construction, generative AI has no concept of truth. The best it can do is synthesise content that seems true to humans, while frequently being false. Yet the biggest, richest tech firms in the world – Google, Microsoft, Facebook, Apple – have already strapped LLMs onto their news and search offerings, with predictably bad results.

// Why?

Why is this happening? If you ask a tech bro, he'll weave you the fantasy that GenAI is freeing humans from the shackles of work, ushering in a utopia where everyone has unlimited leisure time to pursue their dreams, and consigning sickness, death and climate change to history.

Back in reality, literally none of that is happening. Generative AI is only 'good' at the things that humans enjoy doing: painting, writing, making music; creating. (And its 'creations' are problematic, shabby and soulless.) It cannot mop floors or collect rubbish or clean toilets or weed gardens or file taxes (without making dangerous mistakes) or fix climate change.

Billionaries roll out and make huge profits from GenAI not because it's good for humanity, but because it's easy to make huge profits from.

GenAI is a solution to a problem nobody had. We did not need machines to make fake art, and we still don't.

The issue here is not with the technology. Narrow and non-generative AI have enormous beneficial potential.

The issue is with the way GenAI is being trained, marketed, sold and lied about by a small group of extremely rich and unprincipled humans.

When you ask ChatGPT or Gemini for 'help', what you get back is the informational equivalent of junk food: easy, fast, seductive, deceptive short-term gratification that will ultimately leave you worse off than before.

We need and deserve a better diet. Find out more about Writing the Wrongs of AI.