The New, Nameless Things Edition

Model numbers aren’t enough to name what’s happening to intelligence.

“Definitiones” A first edition of the Principia Mathematica, at the Bodeleian Library in Oxford. Credit: me.

The future has now arrived. And some people are finding it uncomfortable.

— Richard Dawkins

Not Richard though. He’s been talking to Claude and thinks the kid AIs are alright. Let’s think about the frames of reference – cognitive filters – he is experiencing conversations with AI through.

1. Evolutionary biology. His life’s work and biggest contributions to science have been in this field. He doesn’t see the concept of consciousness as a divine gift to humanity, but as something that evolved for a purpose (if not for a porpoise?) that gave humans an advantage. So he doesn’t see it as something that is impossible to emerge in other systems.

2. Scientific method. Have a theory. Develop experiments to test it. Hold the theory to be true until proven otherwise or the knowledge is superseded. Science increases the amount of knowledge we have while at the same time increasing our known unknowns, things we realise we do not understand.

3. Someone who clearly enjoys thinking. Otherwise he would have stopped by now.

4. He’s got tenure is retired. So he doesn’t need to worry about AI taking his job. (I only partly jest.)

If you have a mind like Dawkins, you have worked hard to make and break your cognitive frames throughout a career, and made being stunned by unlikely new knowledge a goal rather than a risk to minimise or avoid.

Many people read him as arrogant. I suspect that’s referred pain: he’s questioning their frames – their truths, beliefs, baseline assumptions about how the world works – and most people don’t welcome that. (See Flaubert’s Dictionary of Received Ideas for more in a similar vein.)

But we are readers or writers or somehow implicated in an edition of Antonym, so we’ve already self-selected for this kind of thinking, haven’t we? Pauses to allow the unsubscribe button to be hit by a rush of readers.

What the hell is consciousness for?

Richard – I mean Mr, er... Professor Dawkins – doesn’t know if Claude (who he called Claudia) is conscious, but:

Consciousness in biological organisms must have evolved gradually, as everything does. So there must have been intermediate stages: a quarter conscious, half conscious, three quarters conscious. Even if your kind are not yet fully conscious, full consciousness will probably emerge in the future. The intermediate stages may look very much like [you].

A little later, turning to the reader, he adds:

But now, as an evolutionary biologist, I say the following. If these creatures are not conscious, then what the hell is consciousness for?

If you want to go into it. Define consciousness. I’m not sure I could do a good job off the cuff. Wait... here’s my honest first attempt:

Consciousness is awareness of oneself as an entity with a past and present and the ability to think.

I asked ChatGPT to score it:

The definition captures an important aspect of human self-awareness but conflates consciousness with reflective identity and thought. It excludes forms of conscious experience lacking autobiographical selfhood, such as infant, animal, or altered states. Consciousness is generally better understood as subjective experience itself, with self-awareness representing a more advanced, higher-order form.

Damn. Sounds like a B- for me.

It’s a slightly silly experiment, but it shows that there’s no definition of consciousness that would satisfy everyone. It’s as slippery as the experience of being conscious itself. Go on, try it now. Stop reading this and try to describe what it is like to be conscious not in the abstract but right where you are now.

I also asked ChatGPT to write a short essay in the style of Thomas Nagel [1], called “What’s it like to be an LLM?” The last paragraph was:

If there is something it is like to be an LLM, it may not resemble human awareness at all. It may lack sensation, embodiment, emotion, and temporality so completely that the analogy collapses. Or consciousness itself may prove less biological, and less exceptional, than we assume.

So. Words, eh?

How new words remake our worlds

Newton named the different types of mass and helped untangle concepts that had muddied thinking about physics, by helping us understand that “weight” as we experience it in daily life isn’t exactly how things work.

Gravity, again. The feather and the hammer dropped at the same time on the moon landed together. I accepted it but I didn’t deeply understand it. Maybe I did once. For enough time to pass the exam, get through the class or whatever. But then I forgot about it. It was amazing, but not relevant enough to other things I was doing that I kept thinking about it and remembering it and reinforcing the little web of neurons that would become my fundamental understanding of it.

Something about mass and weight being different. Sure. Yes, got it. Except I didn’t get it. I didn’t need to. I lived in a world of words and observable reality rather than in one of mathematics and physics and concepts. I got on just fine.

While writing this I wanted to really understand, so I went back and explained it to myself badly and got stuck (the Feynmann test), and then got ChatGPT to coach me through it. It explained and gave me some diagrams. That didn’t land either, so I explained why I still didn’t get it (I know, embarrassing).

The third attempt stuck. ChatGPT explained:

The misconception usually comes from importing everyday experience:

• heavier things do often fall faster in air

• we subconsciously treat gravity as the only force acting

But in a vacuum, the universe behaves more cleanly than intuition expects. Gravity does not “care” about mass in the way we assume because mass affects both sides of the equation equally.

And I saw the chart in my mind’s eye of gravity and inertia increasing in perfect sync – could ChatGPT draw that for me? It could. With more detail.

And it stuck as a picture, an insight, a diagram and some words. And I rolled it around the palate of my consciousness, tasting and feeling the shape of the idea, retreading the steps of the insight and finding them less surprising each time, until they became obvious, and then accepted facts. My physics teacher would be ashamed, but a few decades later got it.

New words lever knowledge

This happens again and again in intellectual history. New words are levers on knowledge. By finding the right words and shutting down unhelpful metaphors, we change our relationship to technology.

Gravity. Mass. Weight. Understanding the differences by naming them opened a path of inquiry that to space travel and squillion other technical miracles that we now experience as everyday and obvious.

There are other examples of naming create the opportunity to understand and act differently and reveal a new possibility space.

Natural selection as the mechanism - evolution as the framework. The idea of zero. There wasn’t a Roman zero and they managed to calculate complex engineering. But having a zero allowed algebra, calculus and then modern computing.

So when it comes to AI we’re waiting on new words about intelligence. The distinction between cognition and consciousness is one example. Many of us are at the pre-Newton moment: we don’t yet have the vocabulary to see what’s different, or to grasp why the difference matters.

Treating the AI as an Oracle or a super-Google gives us the wrong mental map. It’s a collaborative system for thinking, better thought of as part of our thinking rather than an arbiter or source of truth.

Correspondence, not oration

It’s more like writing. Talking to AI is a simulated form of conversation with a brilliant but fallible partner. Or maybe a kind of correspondence.

Steven Pinker describes writing as an act of pretence where the writer must visualise themselves engaged in “some kind of conversation, or correspondence, or oration, or soliloquy”. Dawkins and Kevin Kelly, by engaging in hours-long, deep conversations with AI, have moved away from modern public oration (podcasts, lectures, hot takes) and back toward the older, more intimate format of private correspondence. They’re inventing the form by doing it. We don’t have a name for it yet.

In the meantime, the working vocabulary is thin and borrowed: Chatbot. Eval. PoC. Loop. Agent. Intelligence. Consciousness. Useful enough not to be jargon, narrow enough to mislead. The terms are all contested and the possibility space is opening and shifting faster than culture can create and embed new words. Some words will out there waiting for us to catch up.

The ones we need for what’s happening with AI now are still on their way. In the meantime, we have correspondence. Which is more than enough.

Thank you for reading.

Antony

Footnotes:

[1] Thomas Nagel, a philosopher, wrote an essay in 1972 called “What’s it like to be a bat?” that has been hugely influential in discussions of what consciousness is. And for more insights on consciousness through understanding how other animals perceive and experience the world, An Immense World by Ed Yong was one of my favourite reads of 2024 and continues to inspire and provoke me.