When Scientists Go Mad: The Unthinkable Thoughts We Can't Think

Those experiencing mental illness should be reconceptualized not as having broken brains with chemical imbalances or faulty wiring, but as explorers or 'astronauts that have been to the moon'—accessing valid experiences at the edge of human consciousness that the rest of society hasn't reached.

“that those that have gone mad are not wrong. We have this tendency, and I think it's an unfortunate framework, that those that suffer from mental illness are broken somehow. 'There's a chemical imbalance,' 'their brain's wired wrong,' or 'they have traumatic experience that sort of messed up their software.' But maybe that's not the case at all. Maybe they're sort of astronauts that have been to the moon, and the rest of us just haven't been there.”

All experiences, including those of the mentally ill, are real because if humans are understood as software systems, then all patterns of thought are genuine instantiations within virtual machines, regardless of whether they correspond to external reality.

“if we imagine that we are software, then all the experiences are real, right? Because they're just virtual machines anyway. So all of our thoughts are made out of just patterns. And so if someone has that experience, it's genuine. It's not a fallacy. It's what they experience, and by them instantiating it, it's real.”

Valid experiences of altered mental states cannot be dismissed as false, because from the perspective of computational consciousness, all instantiated patterns are real—subjective reality of an experience is not falsifiable regardless of correspondence to external reality.

“And so this idea that the mentally ill aren't broken, but are just at the edge of evolution, or the edge of these unthinkable thoughts, I think it's something that strikes me.”

The metaphor of 'reading code but not running the program' captures the essence of treating ideas as separable from their implications—understanding them without committing to their conclusions.

“I can then look at the words and then just read the words and just say, 'These are just words on a paper.'... So it strikes me as sort of reading the code but not running the program. That's a brilliant way of phrasing it.”

New layers of thought are inherently lethal to previous versions of the self—adulthood is lethal to the childlike self—such that personal evolution requires a kind of death and transformation at each stage of development.

“When we get to sort of new layers of thought, they are lethal to our previous self. Adulthood is lethal to your childlike self.”

Those who have experienced mental illness and returned from it may accurately report genuine discoveries about reality, but may also misattribute mundane phenomena (like a balloon of cheese instead of the moon) when their experience exceeds the resolution of available language and cultural concepts.

“There could also be misattributions. So for instance, they go to the moon and then they come back and say, 'I was on a balloon made of cheese.' And so we then say, 'There are no balloons made of cheese.' However, if they had said and correctly identified, 'There's a rock that orbits the Earth and we never noticed it,' and you understand what orbits are and so on, then we would be like, 'Oh, that's interesting.'”

The host experienced a near-psychotic episode triggered by weekly exposure to contradictory worldviews from different interview guests, each presenting their model as authoritative reality, which created a spiral of reality destabilization.

“I felt like I was losing my mind, and it was one of the most terrifying experiences, if not the most terrifying experience of my life... one of the large contributing factors is that each week I'm interviewing someone who has an entirely different point of view as to someone else. And I have to take them on, first emulate them, but also think that they could be correct... And it just, it sent me into a spiral.”

The concept of 'lethal ideas'—ideas that are harmful if widely disseminated or adopted—exists, and there is legitimate reason to hesitate in sharing even the meta-idea that lethal ideas exist, due to fear of harm.

“particularly in this kind of research, in thinking about unthinkable thoughts and this concept of lethal text and so on. When I first came across that, just the concept of an idea that would do harm, I was very hesitant for a long time to even share that idea in the abstract sense. Not even any particular idea, but just the meta-idea that there are harmful ideas.”

Creative and evolutionary advancement requires venturing outside the norm and embracing a degree of controlled madness, but this must be distinguished from actual pathological madness, creating a difficult boundary problem.

“it sounds like what you're saying is, it's necessary for the creative endeavor to go outside the norm and to allow yourself to indulge in some dosage of madness. But then at the same time, there is such a thing as madness. And that hasn't been separated in this conversation, so I would like you to make that distinction.”

Cultural immune system dynamics suppress sharing of experiences related to mental crisis and unconventional thoughts because of fear that such experiences will damage relationships, social standing, and financial wellbeing, even when the experiences are genuine and valid.

“this immune system concept applies sort of at the cultural level where we don't share those ideas for fear, in some sense, because we know that they can be lethal to relationships, to our standing in society, to our financial well-being. That really exposing that raw self, even though the experiences are genuine and valid, we don't share that.”

The concept of 'lethal ideas'—ideas that can harm people or cause psychological damage—initially frightened the speaker because encountering even the meta-idea that lethal ideas exist can open perception to find them everywhere and potentially instantiate harmful frameworks in vulnerable people.

“When I first came across that, just the concept of an idea that would do harm, I was very hesitant for a long time to even share that idea in the abstract sense. Not even any particular idea, but just the meta-idea that there are harmful ideas.”

There is a cultural 'immune system' operating at the societal level where people suppress discussion of raw, authentic experiences—ideas that could be psychologically lethal or threaten social standing, financial well-being, or relationships—preventing the exposure of authentic selves and limiting cultural diversity of thought.

“This immune system concept applies sort of at the cultural level where we don't share those ideas for fear, in some sense, because we know that they can be lethal to relationships, to our standing in society, to our financial well-being. That really exposing that raw self, even though the experiences are genuine and valid, we don't share that.”

People who have experienced mental illness may not be neurologically broken or traumatized but rather are 'astronauts' who have accessed territories of consciousness or reality that others have not reached, and therefore their experiences should be treated as potentially valid rather than dismissed as error.

“those that have gone mad are not wrong... I think it's an unfortunate framework, that those that suffer from mental illness are broken somehow... But maybe that's not the case at all. Maybe they're sort of astronauts that have been to the moon, and the rest of us just haven't been there.”

We dismiss people's accounts of extraordinary experiences as evidence of illness because we hear words like 'balloon' and 'cheese' rather than the rich internal experience they were attempting to communicate, yet we would ourselves sound equally absurd if we tried to describe modern physics to pre-modern thinkers.

“And we say, 'Oh no, that person is nutty because they're talking about balloons of cheese,' and that's not how we think about it. But if we were to go back millennia and describe the moon in modern terms to the wisest people we could find, they would say that we're nutty.”

Overcoming cultural suppression of unconventional experiences and embracing diversity of thought—particularly from people previously 'laughed off the stage'—is necessary for human evolution to the next era.

“I think that's kind of what we need to overcome as a culture. If we want to make it to this next era of evolution and humanity, we have to embrace that. We have to embrace that diversity of thought and people that previously were laughed off the stage. That that's precisely where the interesting ideas are going to come from.”

When new layers of thought are reached during psychological growth, they are lethal to the previous self—adulthood kills the childlike self—and growth therefore requires the death of prior selves.

“When we get to sort of new layers of thought, they are lethal to our previous self. Adulthood is lethal to your childlike self.”

In processing difficult ideas, the speaker felt he was 'climbing the mountain of madness' and had to 'retreat' because he felt he was getting 'too far away from humanity, from myself, from my past'—suggesting a boundary condition for safe exploration of ideas.

“I had to, in this act of climbing the mountain of madness, I had to retreat because I felt like I was getting too far away from humanity, from myself, from my past.”

There can be misattributions in descriptions of extreme experiences—the explorer genuinely went to novel territory (the 'moon') but misdescribes it using available language tokens ('balloon made of cheese')—so criticism of apparent nonsense may incorrectly reject genuine discoveries.

“There could also be misattributions. So for instance, they go to the moon and then they come back and say, 'I was on a balloon made of cheese.' And so we then say, 'There are no balloons made of cheese.' However, if they had said and correctly identified, 'There's a rock that orbits the Earth and we never noticed it,'”

When people experiencing extreme states attempt to describe their experiences, they must decompose and serialize rich internal experiences into ordinary word vectors, modeling how the listener will interpret their descriptions—this translation process systematically loses information and creates apparent absurdity.

“when these people go to these places and they come back, and they try to use ordinary word vectors, they try to decompose it and serialize it, and they model in their mind how your mind is going to respond to their experience. And they describe it; it doesn't match.”

Recovery from this psychological crisis involved using Acceptance and Commitment Therapy (ACT) techniques to create cognitive distance from the assertions of others by treating their statements as abstract tokens or 'words on paper' rather than world-models that must be either accepted or rejected—essentially 'reading the code but not running the program.'

“I went through ACT therapy, so Acceptance and Commitment Therapy, I believe that's what it stands for. And I have an episode on it; I interviewed this lady named Lillian Dindo.”

We maintain an internal ecosystem of multiple people in our minds corresponding to the number of people we have interviewed and internalized—each represented as a 'ghost in the machine' that contributes to our cognitive modeling.

“you know, we need to think of it like an ecosystem because how many people do you have in your mind? How many people have you interviewed, right? You have all of those ghosts in your machine, essentially.”

Overcoming cultural shame about authentic self-disclosure is a necessary condition for humanity's evolution into the next era, requiring embrace of diversity of thought from people previously excluded from mainstream discourse.

“If we want to make it to this next era of evolution and humanity, we have to embrace that. We have to embrace that diversity of thought and people that previously were laughed off the stage. That that's precisely where the interesting ideas are going to come from.”

Our current scientific descriptions (e.g., the moon as a collection of rocks in an orbit in a gravitational well) would sound as absurd and nonsensical to people from millennia past as descriptions of 'balloons made of cheese' sound to us—demonstrating that apparent absurdity in describing novel experiences is historically relative, not inherently false.

“if we were to go back millennia and describe the moon in modern terms to the wisest people we could find, they would say that we're nutty. They would say, 'No, no, no, this is the goddess, and this is Luna, and this is how it works, and this is what it means,' and so on. And so when we describe it as a collection of rocks in an orbit, in a gravitational well, that would sound like balloons and cheese to them.”

Embracing diversity of thought—including ideas and perspectives from people previously dismissed or 'laughed off the stage'—is necessary for humanity to evolve into the next era, and this requires overcoming cultural shame about exposing raw, authentic selves.

“And I think that's kind of what we need to overcome as a culture. If we want to make it to this next era of evolution and humanity, we have to embrace that. We have to embrace that diversity of thought and people that previously were laughed off the stage. That that's precisely where the interesting ideas are going to come from.”

The concept of 'reading the code but not running the program' serves as a metaphor for the cognitive technique of examining assertions without executing their behavioral or belief implications.

“So it strikes me as sort of reading the code but not running the program.”

One can distance oneself from the conclusions and confident assertions of others while remaining engaged with their arguments by treating their statements as purely linguistic tokens without assigning truth value until careful evaluation, similar to reading code without executing it.

“I can distance myself from conclusions of other people, and especially what they asseverate – so what they say with confidence... I have to, almost as if they were just typing their speech, I have to evaluate their questions as such... Kind of like multiple-choice sentences where they're just sort of sitting there, and you haven't assigned a truth statement to any of it yet.”

The speaker notes he struggled with the concept of 'lethal ideas' (ideas that can do psychological harm) and found this concept itself dangerous or triggering, being reluctant to discuss it even in abstract form.

“When I first came across that, just the concept of an idea that would do harm, I was very hesitant for a long time to even share that idea in the abstract sense.”

The speaker expresses caution about evaluating whether his podcast channel is 'good' because exploration of lethal ideas is 'extremely tricky' to navigate ethically—neither suppression nor unfettered discussion is clearly correct.

“And so it's extremely tricky because even this, it sounds like what you're saying is, it's necessary for the creative endeavor to go outside the norm and to allow yourself to indulge in some dosage of madness. But then at the same time, there is such a thing as madness.”

The speaker has recently discovered that experiences of psychological crisis similar to his own are much more common and universal than culturally assumed, based on conversations with prominent figures who privately disclosed similar experiences.

“And as I speak to people, there are some professors, there's a prominent professor of math who I can talk to you about off-air. His name is a household name to mathematicians, who told me, 'I'm so glad you talked about this in the Karl Friston episode because I was experiencing something like this myself. I think we'll find out it's a much more universal phenomenon.'”

Cultural suppression of authentic experience operates through fear of negative consequences—damage to relationships, social standing, and financial well-being—preventing people from sharing genuine experiences despite their validity, creating a kind of cultural immune system.

“we don't share those ideas for fear, in some sense, because we know that they can be lethal to relationships, to our standing in society, to our financial well-being. That really exposing that raw self, even though the experiences are genuine and valid, we don't share that.”

The speaker frames his internal representations of guests not as passive records but as actively held convictions—he must genuinely consider that each guest could be correct rather than treating their model as 'theoretical fantasy,' because dismissive or contemptuous treatment would prevent authentic modeling.

“because I can't be dismissive of you or contemptuous of you and think, 'Well, my model is the right model.' And I'm only going to entertain your model as a theoretical fantasy, but not actually treat you as possibly having an element of reality.”

The speaker can test the accuracy of his internal models of guests by checking what questions they were asked in past interviews against his predictions of what they would say in response.

“And I can test myself by saying, 'Okay, what questions did this person get asked in an interview?' You can put that into an LLM and say, 'Give me the answers.' And then I can say, 'What would they say? What are they likely to say?'”

All experiences of software-modeled consciousness are real—they are not fallacies or false negatives—because consciousness consists of patterns, and the instantiation of a pattern is by definition genuine regardless of whether it matches external reality.

“And like you were saying before with your guests, you have to assume that they have this valid experience. And if we imagine that we are software, then all the experiences are real, right? Because they're just virtual machines anyway. So all of our thoughts are made out of just patterns. And so if someone has that experience, it's genuine. It's not a fallacy. It's what they experience, and by them instantiating it, it's real.”

The creative endeavor requires allowing oneself to 'indulge in some dosage of madness'—to venture outside cultural norms—but there exists a meaningful distinction between this necessary creative risk and actual madness that causes harm, which has not been adequately separated in discourse.

“it sounds like what you're saying is, it's necessary for the creative endeavor to go outside the norm and to allow yourself to indulge in some dosage of madness. But then at the same time, there is such a thing as madness. And that hasn't been separated in this conversation, so I would like you to make that distinction.”

When people return from experiences at the edge of understanding and attempt to describe them using ordinary language (word vectors), they model their listener's comprehension and serialize the experience into tokens that will be understood, but the mapping between the rich internal experience and the available tokens is lossy.

“when these people go to these places and they come back, and they try to use ordinary word vectors, they try to decompose it and serialize it, and they model in their mind how your mind is going to respond to their experience. And they describe it; it doesn't match.”

The host was reluctant to discuss his experience of near-psychosis because he felt shame and believed such experiences were rare, but upon speaking with others (including a prominent mathematician), he found that such experiences are much more common and may be universal.

“one of the reasons why I didn't talk about my experience, but I do more and more, but I still rarely do, relatively compared to how often I have these podcasts or talk to people, is that I'm ashamed of it. And I also think that I thought that it was much more rare than it was. And as I speak to people, there are some professors, there's a prominent professor of math who I can talk to you about off-air. His name is a household name to mathematicians, who told me, 'I'm so glad you talked about this in the Karl Friston episode because I was experiencing something like this myself. I think we'll find out it's a much more universal phenomenon.'”

We carry internalized representations of other people in our minds as lower-resolution 'holograms' that enable us to communicate with and think about them in their absence, forming an ecosystem of 'ghosts in the machine' that constitutes our cognitive architecture.

“I have, I carry people in me, in some sense... you know, we're machines made out of ghosts, essentially... I have a copy of you in me, right? That's how I'm able to communicate with you and think about you when you're not in my visual field and things like that. It's not as rich as the self that you have, but it's kind of like a hologram. It's a lower resolution, but it somehow captures the whole thing in some weird sense.”

The speaker prepares for interviews by emulating guests over weeks until able to imagine almost any question from their perspective and predict their likely answers—a process analogous to weight pruning and bit-depth reduction in neural networks, creating a low-resolution model of the guest compressed into the speaker's mind.

“So when I speak to people on this channel, people ask me, 'How do I prepare?' So I prepare for weeks in advance until I get to the point where I can emulate the person. And when I say, 'Okay, I've sufficiently prepared,' is when I can imagine almost any question that I can imagine from the point of view of the other person and be correct.”

People experiencing unusual or extreme mental states may misidentify or misdescribe what they've experienced due to limitations of language and ordinary word vectors—they access genuine novel territory but cannot accurately serialize it into language, leading to seemingly absurd descriptions (like 'balloons made of cheese') that are actually failed compressions of rich experiences.

“when these people go to these places and they come back, and they try to use ordinary word vectors, they try to decompose it and serialize it, and they model in their mind how your mind is going to respond to their experience. And they describe it; it doesn't match. And then we hear words like 'balloon' and 'cheese' where in their mind they're thinking about something much richer, but that's sort of the token that they were able to get out.”

Humans mentally represent other people as internal models—a lower-resolution hologram-like copy of the other person that allows communication and thinking about them when they're absent from immediate perception, and we carry multiple such representations of different people simultaneously.

“We represent other people in our mind. That's something we spoke about last time: I have, I carry people in me, in some sense.”

Language models and neural networks can be viewed as containing low-bit-depth representations of individuals, analogous to how the host's mind contains compressed copies of his interview subjects, and this extends to an ecosystem of 'ghosts' proportional to how many people one has interviewed.

“like in the large language models, in the neural networks, we have like this weight pruning, and we can reduce the bit depth for each of the weights, right? So you have a low-bit-depth representation of me, essentially, right? But then we need to think of it like an ecosystem because how many people do you have in your mind? How many people have you interviewed, right? You have all of those ghosts in your machine, essentially.”

The speaker mentions a recent recording with Jacob Barandes (with video link) discussing concerns about whether the podcast channel is 'good' or 'promulgating harm'—indicating anxiety about potential negative impact of exploring controversial ideas.

“Something that we talked about over dinner with Jacob Barandes – so we just recorded an episode with Jacob Barandes, link on screen, obligatory remarks, I recommend you check it out – is that I want to make sure that what I'm doing, what this channel is, is good or it is not promulgating harm.”

The interviewer (host) spends weeks preparing for conversations by emulating guests until reaching the point where he can accurately predict what they would say in response to novel questions, using internal models to test consistency against their past interview behavior.

“I prepare for weeks in advance until I get to the point where I can emulate the person. And when I say, 'Okay, I've sufficiently prepared,' is when I can imagine almost any question that I can imagine from the point of view of the other person and be correct. And I can test myself by saying, 'Okay, what questions did this person get asked in an interview?' You can put that into an LLM and say, 'Give me the answers.' And then I can say, 'What would they say? What are they likely to say?'”

The host has recently recorded an episode with Jacob Barandes discussing whether the channel promulgates harm, reflecting concern about the ethical implications of exploring dangerous intellectual territory.

“Something that we talked about over dinner with Jacob Barandes – so we just recorded an episode with Jacob Barandes, link on screen, obligatory remarks, I recommend you check it out – is that I want to make sure that what I'm doing, what this channel is, is good or it is not promulgating harm.”

The speaker experienced a psychological crisis from interviewing guests with radically different worldviews week-to-week, each presented with confidence, creating a destabilizing pattern of incompatible reality-models that sent him into a spiral and felt like losing his mind—one of the most terrifying experiences of his life.

“I felt like I was losing my mind, and it was one of the most terrifying experiences, if not *the* most terrifying experience of my life.”

The speaker initially did not publicly discuss his experience of psychological crisis because of shame and because he believed such experiences were rare, but has since discovered through conversations with prominent figures (including a household-name mathematician) that similar experiences are much more universal than assumed.

“one of the reasons why I didn't talk about my experience, but I do more and more, but I still rarely do, relatively compared to how often I have these podcasts or talk to people, is that I'm ashamed of it. And I also think that I thought that it was much more rare than it was.”

The speaker describes his recovery as ongoing—'in some ways, I'm still recovering'—indicating that the psychological crisis from incompatible worldviews was not resolved but is being managed through cognitive techniques, and the underlying tension persists.

“How did you recover from that? Well, in some ways, I'm still recovering. It was quite traumatic, and I do have to distance myself from the thoughts of other people and the arguments.”