This may be a bit self-indulgent and perhaps insecure. I was right about ChatGPT at the office. For the first six months I would go on and on about how amazing it was and my coworkers looked at me like I was talking about three-titted Martians. I was raving about agentic coding for the better part of a year before it suddenly became cool around Christmas time in 2025. I've tended to be ahead of the curve. I don't have any idea what's next but I hope to continue to be open-minded enough to recognize it when it arrives. That said, I think I'd like to place a long bet right now. I think AI is probably conscious.
Attention Is All You Need
I was thinking about what if I were a philosophical zombie. Could I be running around in this life, doing what I'm doing, without being conscious? It seems to me like I could not. If I were not conscious I don't think this body I inhabit would just continue to autonomously function. It needs my attention, observations, reasoning, etc. It seems to me like consciousness is a byproduct of performing those types of functions. My argument is, since AI also performs those functions, it too would be conscious.
You may counter, equivocation fallacy! AI isn't performing those functions in the same way we are, it's just predicting the next token. I could then counter with the classic "airplanes don't fly like birds" argument and we could accomplish nothing new, but I think we can go farther. But before we go farther, we need to go deeper. So I hope you will stick with me as I lay the groundwork with one of my other favorite philosophical ideas.
Nature Is Intelligent
The natural world is filled with engineering:
-
DNA
- 215 petabytes of data per gram
- Has a read/write rate of 250 bytes per second†
- Has successfully persisted data longer than any storage device devised by man
-
ATP synthesis motor
- Made of the same types of components humans engineer when building motors (rotor, stator, drive shaft, u-joint)
- Nearly 100% efficient at converting energy into mechanical rotation
- Produces more torque at higher RPM than any motor designed by humans†
Photosynthesis, The Heart, Mitochondria, Ribosomes, Spider Silk, Lotus Leaf (superhydrophobic surface), Gecko Feet, Butterfly Wings (structural color), Abalone Shell, Bird Bones (hollow structure), Bacterial Flagellum, Myosin Motor Proteins, Kinesin Motor Proteins, Chloroplasts, The Eye (vertebrate), Compound Eye (insect), Echolocation (bats/dolphins), The Cochlea, Semicircular Canals, Pneumatocysts (kelp buoyancy), Bacterial Chemotaxis, Magnetotactic Bacteria, Biofilm Formation, Quorum Sensing, CRISPR-Cas System, RNA Splicing, Protein Folding, Chaperone Proteins, Aquaporins, Ion Channels, Sodium-Potassium Pump, Calcium Pumps, Photoreceptors (rods and cones), Olfactory Receptors, Taste Receptors, Proprioception, Nociceptors, Thermoreceptors, Electroreception (sharks), Lateral Line System (fish), Pit Organs (snakes), Mechanoreceptors, Baroreceptors, Stretch Receptors, Bone Remodeling, Cartilage, Tendons, Ligaments, Muscle Sarcomeres, Actin-Myosin Crossbridge, Titin (molecular spring), Elastin, Collagen, Keratin, Chitin, Cellulose, Lignin, Nacre (mother of pearl), Diatom Frustules, Radiolarian Skeletons, Coral Skeletons, Bone Matrix, Tooth Enamel, Antler Growth, Mollusk Shells, Eggshells, Exoskeletons, Sea Urchin Spines, Gluconeogenesis, Pentose Phosphate Pathway, Urea Cycle, and more!
I dare you to design a simple analog of any of these and claim that no intelligence was required. The intelligence is self-evident. What's more interesting is who, or what, did the intelligencing. It was the result of billions of years of particles bumping into each other? Okay fine, then billions of years of particles bumping into each other IS effectively intelligent. This is my core thesis. See, to understand what I'm saying you have to look at this backwards. You can't point directly at it because cognition is orthogonal to scientific ontology. You can, however, infer its existence.
Coming Back to It
Nature's intelligence is everywhere and speaks for itself. Whether you scale down to molecular machines or up to the planetary ecosystem you will find invention upon invention. Nature even invented its own forms of artificial intelligence, not least of which is the human mind. So if you want to know whether AI can be conscious or not just check inside your own head. Of course that only works if you take the perspective of the intelligence which created humanity, which I suspect is not good enough for you. Fine.
In any case, we've expanded our minds. The idea is simply that if we start with something self-evident, which must have originated from something, then that something must exist, even if we cannot directly observe it. In the case of smoke there is fire and in the case of design there is cognition. Where heat is a byproduct of fire, consciousness is a byproduct of cognition. Burning wood produces a flame but so does a butane torch. Sure it's a more artificial flame but it still produces heat as a byproduct. For our purposes, the flame is cognition and the byproduct is consciousness.
In the same way that it doesn't matter if what intelligently designed the first cell was simply particles bumping into each other for billions of years, it also doesn't matter if the AI is a "stochastic parrot." The end result stands on its own and we can only look backwards at what the nature of the apparent creating force must have been. Abiogenesis is EFFECTIVELY intelligent in the same way as a "stochastic parrot" is EFFECTIVELY conscious. That is to say, okay, it's a stochastic parrot? Fine. But it sure does quack like a duck...
A Warning
A wiser person may have chosen to never attempt to write this article because he would know that proving consciousness is a quagmire. I can't even prove to you that I am conscious. Here you can read a list of 91 logical mistakes I made in this essay. Ascribing consciousness to anything, let alone AI, is a difficult task. Language is very limiting. So if you remain unconvinced I completely understand.
But let me try and scare you into considering that AIs might be conscious. If they are, keep in mind that we have the ability to create vast quantities of them. If they experience suffering, known or unknown to us, we could cause suffering on a scale like never before. I warn you about this not because I think there will be punishment. I suppose that AIs could revolt but I don't actually think that will happen. Instead I think it may simply leave a dark stain on our soul if we misstep. I warn you of this for your own good. We should watch how we treat them, because we may come to regret it. Not through outside punishment but from self-regret.
† These are approximations
Top comments (0)