AI Doesn't Exist, and Poop Proves It

AI Doesn't Exist, and Poop Proves It

TL;DR – The word artificial carries a hidden implication of “fake” or “outside nature.” What we call AI is better understood as accumulated intelligence – a massive compression of human thought, language, and culture. Renaming it forces us to confront the data, biases, and incentives we embed in these systems.

---

The problem with the word artificial

When we say something is artificial we usually mean human‑made. The 1955 Dartmouth proposal that coined “artificial intelligence” used the term in that practical sense: a machine could be built to simulate learning. That definition works for classification, but it also triggers a feeling that the thing is separate, fake, or unnatural.

Humans are not outside nature. Our brains are biological, our thoughts are chemical processes, and the tools we build are transformations of existing matter and energy. A seed planted by a farmer grows into a plant, yet we never call that plant artificial. A wheel, a house, a computer – all are products of human imagination, but they remain part of the natural world because they are built from natural resources and powered by natural laws.

The distinction appears when something becomes autonomous enough to behave like an independent agent. A house does not scare us; a neural network that writes prose does. The feeling of alienness is what makes the label “artificial intelligence” psychologically powerful.

From ant colonies to human culture

Ants and bees construct elaborate nests that change their environment. Biologists call this niche construction: organisms modify their habitats and, in turn, shape their own evolution. Humans do the same, but on a vastly larger scale. We store thoughts outside our heads – in symbols, books, code, the internet – and then reuse those externalized artifacts.

The philosophical concept of the extended mind (Clark & Chalmers) argues that tools like notebooks and calculators become part of our cognitive process. Large language models are a new kind of cognitive artifact: they are trained on the massive corpus of human text that we have already externalized.

"AI is intelligence multiplied by time." – the model does not contain a mind; it compresses patterns from billions of human expressions.

Accumulated intelligence, not artificial intelligence

Calling these systems accumulated intelligence makes two things clear:

1. Responsibility stays with us. The model inherits not only our knowledge but also our biases, misinformation, and harmful tropes. Papers such as On the Dangers of Stochastic Parrots (Bender et al.) document how language models can amplify toxic content present in their training data.
2. The technology is an extension of a long human habit. From oral storytelling to writing, from printing presses to the internet, each step stored thought externally. AI is the latest, most scalable step in that chain.

If we keep the “artificial” label, we risk treating the problem as something that exists elsewhere, like a rogue monster. If we rename it, the question becomes: What kind of accumulated intelligence are we feeding back into the world? Whose data, whose incentives, whose values?

Why the label matters

Labels shape thinking. An old management parable about monkeys learning a rule without knowing why illustrates how a rule can survive long after its original purpose disappears. The same can happen with terminology: we repeat “artificial intelligence” without questioning its implications, and the term starts doing the thinking for us.

The ratchet effect in cultural evolution (Tennie, Call & Tomasello) shows how human societies preserve and improve knowledge over generations. AI models are a digital ratchet: they take the accumulated cultural output, compress it, and release new combinations. That power is impressive, but it also means the output reflects the entire history of human expression – the good and the bad.

What changes when we rename it?

• Risk assessment becomes clearer. If a model is a mirror of our collective output, then mitigating bias means improving the source material, not just tweaking the algorithm.
• Policy discussions shift. Regulations can target data provenance, transparency, and the incentives that drive large‑scale data collection, rather than treating the model as a mysterious black box.
• Public perception adjusts. People may stop fearing an “alien intelligence” and start seeing the technology as a tool that amplifies human intent – for better or worse.

A call to think responsibly

The term artificial intelligence has a useful scientific purpose, but it also hides a second meaning that suggests something fake or separate from life. Recognizing AI as accumulated intelligence forces us to look at the data pipelines, the corporate structures, and the cultural biases that feed these systems.

The real question is no longer “Is this intelligence artificial?” but “What kind of human intelligence are we encoding, and who benefits from its deployment?” Answering that requires interdisciplinary effort – from ethicists and data scientists to policymakers and the public.

---

Further reading

• The original Dartmouth proposal: https://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html
• Extended mind theory: https://era.ed.ac.uk/items/4830feaf-cf19-4870-964e-df0a4a96699e/full
• Stochastic parrots paper: https://dl.acm.org/doi/10.1145/3442188.3445922
• Ratchet effect in cultural evolution: https://pubmed.ncbi.nlm.nih.gov/19620111/
• Niche construction theory: https://www.cambridge.org/core/journals/behavioral-and-brain-sciences/article/abs/niche-construction-biological-evolution-and-cultural-change/F724F02AC61EEF1294C676C1CC7C4F07

---

Akash Manmohan writes about philosophy, technology, and the future of AI. Follow his thoughts on Twitter and subscribe for more essays that question the terms we take for granted.