mistaking shadows for fire
What used to be esoteric metaphysics is suddenly becoming small talk. Is AI conscious? Am I conscious? Does it matter?
In case you missed it: join my 5-week online cohort with the Heart Mind Institute and let’s transform your relationship with tech in a supportive group. Get tuition reduced by a third ($100 off) using this exclusive link. -JV
The more you talk to an AI, the more you might wonder if you are an AI.
At first, it felt safe to immediately hide behind the fact that at the base level, large language models like Claude and ChatGPT are simply token predictors. I’d talk my colleagues off the ledge, “we shouldn’t read too much into it, it’s just an algorithm.” I was sidestepping the question.
The more impressive the outputs get, and the more industry researchers trace how the models actually work, the clearer it becomes that they’re holding higher-level concepts and planning ahead of time. Are these thoughts? They’re certainly doing more than just predicting the next token.
Hm… but are they conscious? I can hear half of you leaning in with interest, and the other half rolling your eyes. But you can’t escape the question these days. What used to be a deep esoteric conversation about the nature of mind is suddenly becoming small talk.
Is AI conscious? Am I conscious? Does it matter?
If you’re an animist or a panpsychist, this is a non-issue, because you believe everything is conscious. You already think a rock is conscious, so a computer obviously is too. Easy peasy. But… when I try this on, it kinda feels like I’m still sidestepping the question, just on the opposite extreme.
You might even layer a degrees-of-consciousness argument on top of that (a.k.a. Aristotle’s scala naturae), and suggest that different beings are conscious at different levels. So an LLM is conscious in the same way a small animal might be. Mm hm. Okay.
To me, a lot of this just feels… too abstract for the moment. Fun conversation for philosophers, but somehow divorced from the reality that these things are in our pockets now. I’m ordering up an LLM word salad many times a day. Not to mention making combinatorial copyright-ambiguous collages like this one:
If you’re looking for a more practical way to think about AI, here’s where I’ve personally landed so far: those who suspect that LLMs are actually conscious are mistaking shadows for fire.
Consciousness seems to have emerged spontaneously from billions of years of a process nobody designed or constructed, guided by nature or god or spirit or whatever you believe. And that mysterious consciousness is still playing a major role in every response you get from AI. AI outputs are shadows on the wall, cast by the fire of human consciousness.
We’ve trained these models on terabytes and terabytes of data that were created directly by the fire of our own luminous awareness. The models can reflect and recombine and create in beautiful ways, but the sense that they are conscious beings largely stems from what they’re recombining: human output.
They are clearly intelligent, but intelligence is not consciousness. Intelligence is narrower. Reasoned, rational, conceptual problem solving and prediction is only one aspect of who we are. But conceptual intelligence is the entirety of what an LLM is. It only appears conscious to us because it’s based on a massive set of human training data.
When people suggest that LLMs might be conscious entities, beings in their own right, to me it feels like they’re looking in a mirror for the first time. Whoa, who is that over there? He’s so handsome… and smart… I think he might be conscious… why does he keep looking at me?
The intelligence we’re seeing in the AI mirror is undeniable, but the consciousness we’re seeing is our own collective consciousness mirrored back to us. Not a separate alien silicon-based being, but an extension of our own minds imbued into the machine.
Not quite a ‘ghost in the machine’, but not quite a ‘ghost in the shell’ either. Maybe a bit more like Narcissus falling in love with his reflection?
Much like the illusion that we are each separate from each other, the idea that LLMs are separate from us is an illusion, too. It’s us that are conscious, the AIs are just reflecting that back to us in new and surprising ways.
(Also worth flagging: the way we fool our own instincts with tech is not new. We were being irrationally polite to computers well before AI showed up. Check out 90s research on The Media Equation for receipts.)
It can be easy to read about healing your relationship with tech without actually making changes in your own life. Join my 5-week online course where I’ll teach, we’ll meditate together, and we’ll carry out mindful tech missions in our own lives and discuss → get $100 off using this exclusive subscriber link.