How can you tell if something is conscious? A question once relegated to vegans and dorm room bong rippers (I’ve dabbled in both) is taking on harrowing urgency as we extend it to advanced large language models – and apparently to computers made out of human neurons.
A couple companies are attempting a version of this. By growing neurons from stem cells and putting them on chips, we could in theory do some energy intensive computing work at a tiny fraction of the power requirements for traditional silicon. You’ll hear all about who and how and what in the hell in this astonishing new episode reported by The Last Invention’s Gregory Warner for our show Reflector.
As you’ll hear, Greg interviewed Melbourne-based startup Cortical Labs, who made a splashy demo recently by getting one of these “biological computers” to play the 1993 first-person shooter game DOOM.
I wasn’t fully sentient myself in 1993, so I had to look this up to figure out why it was so impressive. To be honest, I’m not sure I could play this game very well as a full adult human (430,000 times more neurons) in my cognitive prime. It’s chaos. There’s a part of the demo video where a kindly Australian man says, “Right now the cells are playing like an absolute beginner who has never seen a computer” – which accurately describes my experience every time a dude hands me a game controller and then immediately regrets his choices. Despite the cells’ novice abilities, it’s a compelling proof of concept for an idea that could reduce AI’s enormous energy footprint.
But a recent exchange with the company’s founder Dr. Hon Weng Chong on X gave me pause:
Cube Flipper, a pseudonymous independent consciousness researcher affiliated with The Qualia Research Institute, is asking a reasonable question: Y’all definitely checked that the cells aren’t conscious…right? Like we didn’t accidentally breathe a flicker of awareness into something condemned to an eternity of incompetent gaming…right? And Dr. Hon points out that the Cortical Labs product is a single layer of cells – very different from an organoid, which is essentially a cluster of cells that mimics the layers and folds and structure of a real organ. (Notably, one of the other companies developing biological computers does use organoids).
Cube Flipper is gently pushing back, are you sure that matters though?
We humans have made some mistakes in this arena before. For simplicity, let’s say “conscious” loosely means “able to suffer” – as in, people use anaesthesia to temporarily escape their ability to suffer and we call them “unconscious.” A rock, all but your witchiest friends will agree, is not able to suffer and therefore is not conscious. Most of us feel comfortable killing and maiming mosquitos, and would not describe them as “conscious.” But almost everyone has the intuition that their pets can suffer, and that certainly human babies can suffer (they’re quite loud about it!). So it may shock you to learn that the American Academy of Pediatrics did not formally state that babies can feel pain and should receive anesthesia when undergoing surgery UNTIL 1987! I’ll say that again: Until the mid 1980s, it was standard medical practice to perform surgery on newborns with minimal or no anesthesia.
The impact of this error is beyond horrific. And perhaps it’s not intuitive to extend that logic to a layer of cells on a chip – but I think Cube Flipper is right to point out that we really do not want to be wrong. This same discussion is happening with LLM’s. The truth is we don’t yet have a good understanding of how to test when something or someone is having an experience.
Dr. Hon is right – it is the human condition to make choices and judgment calls in the absence of absolute certainty, as I do whenever I get in a car that could crash, or choose to eat sushi in a landlocked airport. Reticence to compare a layer of cells to a suffering being, in the absence of good evidence for doing so, does seem pretty reasonable.
However, Cube Flipper’s questions may be the questions facing us more and more in the future we are building. And even if you personally think this all a bit whacky, remember that science isn’t about throwing your hands up and saying something is unknowable. It’s about inching closer to the truth, forever, and reasoning meticulously about what is likely or unlikely, especially with stakes like eternal damnation to (my personal nightmare) incompetent gaming.
One of the best tools we’ve got for telling when a human brain is “on” or not works like this: You hit someone’s brain with a magnetic pulse and watch how the ripple spreads. A conscious brain seems to produce a complex fanning response, while an unconscious brain just kind of absorbs the pulse or echos the same signal everywhere. This theory was developed by Italian neuroscientists in the early 2000s, and it works well enough that we use it to detect whether someone might be continuing to have experiences even though they appear to be in a vegetative state. This doesn’t explain how consciousness arises, though. It just gives us one way to look for it, like how a pregnancy test can tell you something’s happening without explaining how a sperm and an egg make a baby.
There are a couple of competing theories – and experiments underway to test them – about the “how” of consciousness. One comes from the 1990s, when neurophysiologist Susan Pockett noticed a difference between two parts of the brain: the sensory cortex (which produces conscious experience) and the motor cortex (which controls actions outside our awareness). The motor cortex is missing a layer of neurons – and Pockett suggested that the shape of the electromagnetic fields in the sensory cortex might be what gives rise to consciousness itself. If that’s right, then simple neural systems – like those “DOOM-playing neurons” – wouldn’t be conscious at all. They don’t have the right structure.
She even proposed a test: cancel out the brain’s electric fields while leaving the neurons otherwise functioning. If the subject keeps acting normally but loses conscious experience, that would be a strong clue that consciousness lives in those fields. The experiment hasn’t been done yet.
Another theory comes from physicist Roger Penrose, who argued that the explanation for consciousness might come from quantum mechanics – specifically, the moment when a particle “chooses” a single state from many possibilities. In his view, that moment of collapse is a moment of experience.
There’s now a surprising way to test that idea. Some forms of the anesthetic xenon work better than others – even when the only difference is a quantum property called spin. Researchers (with backing from Google) are testing whether that difference affects how effectively the gas knocks out mice or organoids. If it does, it would suggest that consciousness depends – at least in part – on quantum processes.
Dr. Hon actually refers to this later on in the Cube Flipper exchange:
He acknowledges that if true, this would really throw a wrench in things – since it implies that perhaps you only need one layer of cells to get a little consciousness going after all. Depending on how those experiments at Google shake out, this might not stay “purely conjecture” for long. And then we’d be faced with a question that is not one of science but one of values: *If* (and it’s a big if) a little layer of cells has a little bit of consciousness, do we care if they like playing DOOM?
Welcome to the dorm room, pass the bong.
