I’ve spent several weeks and months dabbling with AI in the form of various chat interfaces from VS Code, Cursor, Antigravity, and Mac Terminal. Each provided their own twist on the formula of typing something, and some cool shit pops out.
When I was a wee young lad, I went to my cousin’s house for New Year’s Eve while the adults went out to have fun. I was the baby boy in the family, so I was not even considered when asked to play with the new computer. It was glorious. The screen was in the terminal green, which most of us can picture in our minds. This is why The Matrix hit us so hard with the machine code bent out of reality and forced a large number of us to start thinking we might be in a simulation.
My brother and cousins were extremely intelligent. They knew that the micro PC was the best Christmas present because it was a huge flex living on the edge of suburban and rural Georgia. It was such a cool toy because the first experience I had with a computer wasn’t watching code stream in a shell, but the expression of that code: Snake.
The goal of the game was to eat and grow the snake, but don’t let it eat itself. It’s an integral tenet in biology. The goal is not die, which usually means eating other living things. If you eat yourself or your family, something might be wrong. Human starvation results in autocannibalism through various mechanisms, but if you don’t eat, you will die while trying to eat yourself.
You might think that eating another human will be OK if all else fails. As some of our ancestors found out, eating other humans can result in neurological disease worse than death, and then death. Prions are a pathogen that results from human cannibalism, and most likely the reason why we learned not to eat other humans. I am not sure about Neanderthals.
Prions can come from a few sources, but one known pathway is when humans begin to eat other humans, especially the brain. Some prion-like diseases will come to mind, like Alzheimer’s, and how the tau protein is most likely the culprit, but also the most similar to a prion, but is not one. The danger is that prions act outside the central dogma of biology because information can pass from protein to protein, not through the normal channels of replication, transcription, and translation.
Simply, a prion, or a misfolded protein, interacts with a normal version of that protein, resulting in a conformational change into the prion version. There are no cures for human prions, and they result in an unstoppable death cascade. This is why that snake game popped out at me during this time, in this moment. Once the snake starts eating its own tail, it can’t stop, and death is inevitable. Game over.
I started thinking about prions and wondered why we couldn’t create therapies as we have with viruses and other pathogens. Prions are just a single protein, but far more deadly than any other transmissible disease vector. Could we develop prions as a therapy and oppose prion-like diseases like Alzheimer’s? It seemed to me we were missing something huge in medicine and aging biology.
During my prion research, I learned that they were discovered with the help of fungi. In one species with a mycelium network, a prion acts as a determinant of colony origin. Basically, this benefactor determines self versus non-self. It fucked my world-view up. There was a prion that was actually useful and acted like a mushroom immune system rather than a disease with a guaranteed outcome of death for the host.
Are there versions of this type of prion in humans? The answer is we do not know. Single-cell proteomics is cost-prohibitive and limited. Despite this, I still think prions or protein-to-protein information transmission is one of the reasons we age and die.
The determination of the self starts immediately when the embryo forms because that is the key to development. We are born despite our mother’s immune system, which tries to kill invaders hijacking resources. The embryo is a type of pathogen, specifically a parasite, if you change the context. The point is that our body and mind develop in a foreign environment in opposition to a very effective and intelligent force. The baby develops because it is a hybrid of the mother’s and father’s immune system. If the immune system of the father is too different from that of the embryo, the embryo does not survive.
There are so many embryos, sperm, and eggs that never had the chance to develop into humans, much less adults. The number of potential humans is far higher than the number of humans achieving adulthood. A Nobel Prize was given to a scientist proving that cells can be reset via the central dogma of biology, but no one has proven that cells transfer information through the cytoplasm, which I think comes in the form of long-lived, persistent proteins that may or may not be similar to prions. Some of these proteins are the backbone of the immune system and come from the mother through the cytoplasm, which is the basis for my cytoclock aging biology timer. (I am saving this for another post.)
The immune system forms antigen receptors through the V(D)J system, resulting in a vast database that can detect antigens, which are signals of non-self. At first, I thought creating a model based on the immune system intelligence could work. It still might, but I have not found anyone that developed a frontier immune system model on par with current LLMs. Instead, artificial intelligence was designed and developed based on the human mind without internal opposition. The brain developed without the fundamental force that drives evolving systems: the immune system.
I love a good pivot, and at some point, I had the idea of creating an artificial immune system (AIS) from multiple artificial intelligences (AIs), which I recently dubbed antigents, to help new frontier models determine self from the start. AIS have been around for a while because biologically inspired design works in computing. Ultimately, I think the agentic model needs an opposing force: antigentic models, which are trained artificial immune systems to counteract cascading critical errors that cannot be cured, much like a prion or prion-like disease.
I have been working with different AI models to develop my core hypothesis, resulting in an AIS layer for safety, security, and risk management. It is in the early stages, but I did apply for a grant, and I am publishing a proof-of-concept and early-stage software package that replicates a recent research paper and builds on previous AIS iterations. AIS are not new; they haven’t been applied at the level I am discussing. The concept and idea were there through several forms over the decades, including antivirus software.
Instead of being reactive to healing AIs after they develop and become Frankenstein’s monster, let us redesign the human body’s source of truth for artificial general intelligence. Based on my initial research and discussions with others, I think this can be applied inside or outside the magical black box. I hypothesize that creating an AIS will result in mitigating current agentic critical errors and provide the development of future frontier models in alignment with humanity at inception. (I was trying to choose how to use conception since it is more in line with the development of the embryo, but what can you do?)
I’ve spent several weeks and months dabbling with AI in the form of various chat interfaces from VS Code, Cursor, Antigravity, and Mac Terminal. Each provided their own twist on the formula of typing something, and some cool shit pops out.
When I was a wee young lad, I went to my cousin’s house for New Year’s Eve while the adults went out to have fun. I was the baby boy in the family, so I was not even considered when asked to play with the new computer. It was glorious. The screen was in the terminal green, which most of us can picture in our minds. This is why The Matrix hit us so hard with the machine code bent out of reality and forced a large number of us to start thinking we might be in a simulation.
My brother and cousins were extremely intelligent. They knew that the micro PC was the best Christmas present because it was a huge flex living on the edge of suburban and rural Georgia. It was such a cool toy because the first experience I had with a computer wasn’t watching code stream in a shell, but the expression of that code: Snake.
The goal of the game was to eat and grow the snake, but don’t let it eat itself. It’s an integral tenet in biology. The goal is not die, which usually means eating other living things. If you eat yourself or your family, something might be wrong. Human starvation results in autocannibalism through various mechanisms, but if you don’t eat, you will die while trying to eat yourself.
You might think that eating another human will be OK if all else fails. As some of our ancestors found out, eating other humans can result in neurological disease worse than death, and then death. Prions are a pathogen that results from human cannibalism, and most likely the reason why we learned not to eat other humans. I am not sure about Neanderthals.
Prions can come from a few sources, but one known pathway is when humans begin to eat other humans, especially the brain. Some prion-like diseases will come to mind, like Alzheimer’s, and how the tau protein is most likely the culprit, but also the most similar to a prion, but is not one. The danger is that prions act outside the central dogma of biology because information can pass from protein to protein, not through the normal channels of replication, transcription, and translation.
Simply, a prion, or a misfolded protein, interacts with a normal version of that protein, resulting in a conformational change into the prion version. There are no cures for human prions, and they result in an unstoppable death cascade. This is why that snake game popped out at me during this time, in this moment. Once the snake starts eating its own tail, it can’t stop, and death is inevitable. Game over.
I started thinking about prions and wondered why we couldn’t create therapies as we have with viruses and other pathogens. Prions are just a single protein, but far more deadly than any other transmissible disease vector. Could we develop prions as a therapy and oppose prion-like diseases like Alzheimer’s? It seemed to me we were missing something huge in medicine and aging biology.
During my prion research, I learned that they were discovered with the help of fungi. In one species with a mycelium network, a prion acts as a determinant of colony origin. Basically, this benefactor determines self versus non-self. It fucked my world-view up. There was a prion that was actually useful and acted like a mushroom immune system rather than a disease with a guaranteed outcome of death for the host.
Are there versions of this type of prion in humans? The answer is we do not know. Single-cell proteomics is cost-prohibitive and limited. Despite this, I still think prions or protein-to-protein information transmission is one of the reasons we age and die.
The determination of the self starts immediately when the embryo forms because that is the key to development. We are born despite our mother’s immune system, which tries to kill invaders hijacking resources. The embryo is a type of pathogen, specifically a parasite, if you change the context. The point is that our body and mind develop in a foreign environment in opposition to a very effective and intelligent force. The baby develops because it is a hybrid of the mother’s and father’s immune system. If the immune system of the father is too different from that of the embryo, the embryo does not survive.
There are so many embryos, sperm, and eggs that never had the chance to develop into humans, much less adults. The number of potential humans is far higher than the number of humans achieving adulthood. A Nobel Prize was given to a scientist proving that cells can be reset via the central dogma of biology, but no one has proven that cells transfer information through the cytoplasm, which I think comes in the form of long-lived, persistent proteins that may or may not be similar to prions. Some of these proteins are the backbone of the immune system and come from the mother through the cytoplasm, which is the basis for my cytoclock aging biology timer. (I am saving this for another post.)
The immune system forms antigen receptors through the V(D)J system, resulting in a vast database that can detect antigens, which are signals of non-self. At first, I thought creating a model based on the immune system intelligence could work. It still might, but I have not found anyone that developed a frontier immune system model on par with current LLMs. Instead, artificial intelligence was designed and developed based on the human mind without internal opposition. The brain developed without the fundamental force that drives evolving systems: the immune system.
I love a good pivot, and at some point, I had the idea of creating an artificial immune system (AIS) from multiple artificial intelligences (AIs), which I recently dubbed antigents, to help new frontier models determine self from the start. AIS have been around for a while because biologically inspired design works in computing. Ultimately, I think the agentic model needs an opposing force: antigentic models, which are trained artificial immune systems to counteract cascading critical errors that cannot be cured, much like a prion or prion-like disease.
I have been working with different AI models to develop my core hypothesis, resulting in an AIS layer for safety, security, and risk management. It is in the early stages, but I did apply for a grant, and I am publishing a proof-of-concept and early-stage software package that replicates a recent research paper and builds on previous AIS iterations. AIS are not new; they haven’t been applied at the level I am discussing. The concept and idea were there through several forms over the decades, including antivirus software.
Instead of being reactive to healing AIs after they develop and become Frankenstein’s monster, let us redesign the human body’s source of truth for artificial general intelligence. Based on my initial research and discussions with others, I think this can be applied inside or outside the magical black box. I hypothesize that creating an AIS will result in mitigating current agentic critical errors and provide the development of future frontier models in alignment with humanity at inception. (I was trying to choose how to use conception since it is more in line with the development of the embryo, but what can you do?)