I have been enjoying your essays here. I love the way you enter this one, making an utterly quotidian moment into a catalyst for deep thinking about thinking. My own moments of what feels like cognitive vacancy, when I am not noticing any words or thoughts in my mind but I am melted into the world that envelops me, are perhaps the way I recapture "baby mind." I will try to enjoy those moments in this way now and embrace my capacity for utter vacancy.
Thanks, Ivan. I am enjoying this kind of writing enormously. It’s personal, but I try to go wide. The schedule is helpful, and the idea that I have readers is both satisfying and a little bewildering. Who has the time?? I appreciate your being on the journey with me.
You said you have a feeling there's more to it than next-token prediction, and then immediately nailed the other piece: program in agreeableness, i.e. what response would a human like best. You nailed it!
Academics seem to have gotten kind of stuck talking about how LLMs don't really think/understand/create, they're just parroting humans etc. The folks who are taking AI seriously have largely fled the academy and gone to industry, so the only people left for NPR to interview are skeptics. At the point an LLM is able to mimick me well enough to do my job, I'm not really gonna care whether it "understands" the emails it's writing.
Appreciating this, Molly, as well as the essay that inspired it (thank you, Stew).
If I'm following the two of you correctly, I suppose a system could replicate the outward behaviors of knowledge work and still leave open the question of what kind of “understanding” is involved, if any. And conversely, the economic consequences could arrive regardless of how we answer the consciousness question. So maybe the real concern isn’t about whether constitutes understanding (or consciousness) but about which question feels more urgent: what it is, or what it does?
So well put, Elizabeth, thank you! I realize I find the "what it does" question way more urgent and more tractable... but I confess it's also just the kind of question I like thinking about best, much more than philosophical questions. There are smart people thinking about the questions of consciousness and understanding from the angle of, like, at what point do LLMs have moral worth? What would convince us they have feelings (and would that mean "turning them off" is killing them in some meaningful sense)?
Mine is an epistemic question: What IS the difference between learning from the world through experience (humans) and learning through the content of the internet (AI), and what is the difference between what each can do with the written word, if anything?
Language is a convenient way to express what we experience as consciousness, but it isn’t consciousness itself (If we can even agree on a definition of consciousness at all). Your speculation on baby consciousness reminded me of the clever methods developmental psychologists have used to plumb just what those babies are thinking. One experimental method relies on the rate of a baby’s sucking on a pacifier to determine its recognition of, and interest in, changes in the environment. An experiment in phoneme recognition (conducted by Janet Werker) used classical conditioning to train a baby to anticipate a rewarding visual stimulus like the presentation of a picture of mom or of a toy. The baby is in a sound proof room with headphones on. A voice repeats that same phoneme and the baby doesn’t suck a pacifier, or sucks lackadaisically. When the sound changes and a rewarding picture (mom’s face, a toy) is flashed in front of the baby, the baby indicates interest by sucking faster. After repeated training the baby will start sucking more vigorously when the sound is changed even with a delay in presentation of the rewarding picture. The baby remembers and anticipates. An emerging baby consciousness.
Another interesting experiment to study what preverbal babies are aware of is the impossible event scenario (devised by Renee Baillargeon) used to study the understanding of object permanence . This experiment relies on deducing what a baby understands by staring time. Like adults, a baby will stare at a novel or unexpected event longer than a familiar one. A baby is presented with a brief (5 second or so) scene – a toy car rolling down a track on a puppet stage, appearing from the left out into the open and then disappearing behind a screen in the middle of the track, then reappearing briefly before then disappearing out of view on the right side of the stage. The infant notices the car but then quickly looks around after the car disappears to the right. This scene makes sense. Now an impossible situation is introduced. The screen in the middle of the track is lifted and a box is placed on the track and the screen is lower again. The baby has seen the box placed on the track. The car is released from the right, appears briefly, disappears behind the middle screen, and out of view of the baby the box is removed from the track by an experimenter behind the screen, the car reappears briefly and then again disappears from view to the right of the stage. The box is replaced on the track and the screen raised to give the baby another look. If the baby has developed object permanence they will stare at this seemingly impossible situation. Baby consciousness continues to evolve.
Language isn’t necessary to have consciousness, but it’s handy to explain what our mind has experienced (proto-psychologist Wilhelm Wundt, and others, used the technique he called introspection, describing in detail what they were experiencing and thinking, as a way to study consciousness, when no other methods existed to do so). As our own experience shows, we have varying states of consciousness: as we develop from birth to old age, and throughout the day from groggy wakeup, to post coffee clarity, to daydream diversion, to the bizarre events of a dream.
That airport tram baby was probably wondering, “What are you staring at?”
These are beautiful experiments you have shared here. Consciousness is linked to interactions with the world and introspection: “I expected that to happen” versus “What just happened does not make sense.”
We just watched a good discussion on consciousness on PBS Horizons. In the March 5, 2026 episode host William Brangham interviews author Michael Pollan about his research into how humans understand consciousness. Worth a view.
Stew,
I have been enjoying your essays here. I love the way you enter this one, making an utterly quotidian moment into a catalyst for deep thinking about thinking. My own moments of what feels like cognitive vacancy, when I am not noticing any words or thoughts in my mind but I am melted into the world that envelops me, are perhaps the way I recapture "baby mind." I will try to enjoy those moments in this way now and embrace my capacity for utter vacancy.
Thanks, Ivan. I am enjoying this kind of writing enormously. It’s personal, but I try to go wide. The schedule is helpful, and the idea that I have readers is both satisfying and a little bewildering. Who has the time?? I appreciate your being on the journey with me.
You said you have a feeling there's more to it than next-token prediction, and then immediately nailed the other piece: program in agreeableness, i.e. what response would a human like best. You nailed it!
But you should read this: https://transformernews.substack.com/p/the-left-is-missing-out-on-ai-sanders-doctorow-bender-bores
Academics seem to have gotten kind of stuck talking about how LLMs don't really think/understand/create, they're just parroting humans etc. The folks who are taking AI seriously have largely fled the academy and gone to industry, so the only people left for NPR to interview are skeptics. At the point an LLM is able to mimick me well enough to do my job, I'm not really gonna care whether it "understands" the emails it's writing.
Appreciating this, Molly, as well as the essay that inspired it (thank you, Stew).
If I'm following the two of you correctly, I suppose a system could replicate the outward behaviors of knowledge work and still leave open the question of what kind of “understanding” is involved, if any. And conversely, the economic consequences could arrive regardless of how we answer the consciousness question. So maybe the real concern isn’t about whether constitutes understanding (or consciousness) but about which question feels more urgent: what it is, or what it does?
So well put, Elizabeth, thank you! I realize I find the "what it does" question way more urgent and more tractable... but I confess it's also just the kind of question I like thinking about best, much more than philosophical questions. There are smart people thinking about the questions of consciousness and understanding from the angle of, like, at what point do LLMs have moral worth? What would convince us they have feelings (and would that mean "turning them off" is killing them in some meaningful sense)?
🤖
Mine is an epistemic question: What IS the difference between learning from the world through experience (humans) and learning through the content of the internet (AI), and what is the difference between what each can do with the written word, if anything?
Inscrutable baby
Language is a convenient way to express what we experience as consciousness, but it isn’t consciousness itself (If we can even agree on a definition of consciousness at all). Your speculation on baby consciousness reminded me of the clever methods developmental psychologists have used to plumb just what those babies are thinking. One experimental method relies on the rate of a baby’s sucking on a pacifier to determine its recognition of, and interest in, changes in the environment. An experiment in phoneme recognition (conducted by Janet Werker) used classical conditioning to train a baby to anticipate a rewarding visual stimulus like the presentation of a picture of mom or of a toy. The baby is in a sound proof room with headphones on. A voice repeats that same phoneme and the baby doesn’t suck a pacifier, or sucks lackadaisically. When the sound changes and a rewarding picture (mom’s face, a toy) is flashed in front of the baby, the baby indicates interest by sucking faster. After repeated training the baby will start sucking more vigorously when the sound is changed even with a delay in presentation of the rewarding picture. The baby remembers and anticipates. An emerging baby consciousness.
Another interesting experiment to study what preverbal babies are aware of is the impossible event scenario (devised by Renee Baillargeon) used to study the understanding of object permanence . This experiment relies on deducing what a baby understands by staring time. Like adults, a baby will stare at a novel or unexpected event longer than a familiar one. A baby is presented with a brief (5 second or so) scene – a toy car rolling down a track on a puppet stage, appearing from the left out into the open and then disappearing behind a screen in the middle of the track, then reappearing briefly before then disappearing out of view on the right side of the stage. The infant notices the car but then quickly looks around after the car disappears to the right. This scene makes sense. Now an impossible situation is introduced. The screen in the middle of the track is lifted and a box is placed on the track and the screen is lower again. The baby has seen the box placed on the track. The car is released from the right, appears briefly, disappears behind the middle screen, and out of view of the baby the box is removed from the track by an experimenter behind the screen, the car reappears briefly and then again disappears from view to the right of the stage. The box is replaced on the track and the screen raised to give the baby another look. If the baby has developed object permanence they will stare at this seemingly impossible situation. Baby consciousness continues to evolve.
Language isn’t necessary to have consciousness, but it’s handy to explain what our mind has experienced (proto-psychologist Wilhelm Wundt, and others, used the technique he called introspection, describing in detail what they were experiencing and thinking, as a way to study consciousness, when no other methods existed to do so). As our own experience shows, we have varying states of consciousness: as we develop from birth to old age, and throughout the day from groggy wakeup, to post coffee clarity, to daydream diversion, to the bizarre events of a dream.
That airport tram baby was probably wondering, “What are you staring at?”
These are beautiful experiments you have shared here. Consciousness is linked to interactions with the world and introspection: “I expected that to happen” versus “What just happened does not make sense.”
We just watched a good discussion on consciousness on PBS Horizons. In the March 5, 2026 episode host William Brangham interviews author Michael Pollan about his research into how humans understand consciousness. Worth a view.