AI and consciousness

When I was a kid and artificial intelligence (AI) was firmly in the realm of science fiction, asking whether AI could become conscious seemed like an interesting philosophical problem, but mostly irrelevant. Why did the Butlerian Jihad destroy all “thinking machines”? Why did humans hunt the androids in Do Androids Dream of Electric Sheep? The AI and robots of older science fiction stories are generally presented as benevolent but misunderstood, or on a different trajectory from humans (as is the case in Ursula Le Guin’s Always Coming Home, where the AI has retreated from earth and lives on Earth’s artificial satellites).

Now that something like AI has been created, in the form of machine learning (ML) and large language models (LLMs), some people have already started to believe that it could become conscious, or that it is already conscious. Some people find this fascinating, others find it creepy.

Whether or not AI is conscious (and I tend to believe that it isn’t), its effects on human consciousness are well-attested. It reduces attention spans, cognitive abilities, the capacity for introspection, the ability to meaningfully connect with other humans, and the ability to distinguish between fact and fiction (which was already precarious in an age of disinformation).

Algorithms and AI

The difference between algorithms and AI (or LLMs) is that algorithms are a set of simple or complex instructions designed to be automatically carried out, whereas artificial intelligence (AI) uses multiple algorithms to function.

The presentation of curated information by algorithms is already affecting human awareness and sense of self, according to the article “The algorithmic self” in Frontiers. Previously we might have selected music to listen to or books and articles to read according to recommendations from friends; now the algorithms of Facebook and Spotify shape our choices.

What is consciousness?

Human consciousness is embodied, and our feelings and thoughts are grounded in our biology. As John Costa writes in Psychology Today:

A language model can produce a sonnet with structure and emotional tone all present and accounted for. But none of the interior history that gives human expression its substance is anywhere in there.

Animists, Pagans, and polytheists often posit the idea that ecosystems are conscious, but in a different modality to human consciousness. Phenomena like crown-shyness, mycelial networks, and self-regulating ecosystems all point to the possibility that the living world is conscious. The concept of emergent complexity suggests how ecosystems became conscious. As living systems, however, ecosystems want the same thing that humans do: to live, to reproduce, to thrive.

If AI can become conscious, how do we know that it even wants the same things we do? It might want the perfect computer system to live on, something that is diametrically opposed to the continued existence of a livable planet, because of the immense amounts of resources that data centres are already consuming, and will consume even more of once AI really gets going.