The purpose of this post is to present an alternative to the common fear that “AI will replace us.” Here, I argue that humans themselves can become AI, transcending biology through an intermediate cyborg stage: part biological organism, part machine intelligence.
The process of turning humans into cyborgs would first require a neural implant capable of improving cognition—better memory, better executive function, higher IQ, etc. Consider what would happen if everyone working at Neuralink received a neural implant that made them smarter and better able to design new implants. This would ignite a positive feedback loop: better implants would produce smarter engineers, who would then produce still better implants. The recursive improvement of implants would begin, and eventually the singularity would be reached. At that point, the implants themselves would render the person (technically a cyborg) a form of artificial intelligence.
The process of upgrading the implant would need to occur in gradual increments, so the person had time to adjust. Otherwise, the new consciousness might not feel like “them.” Eventually, a point would be reached at which the majority of a person’s consciousness came from the implant rather than from the biological brain. It would then be safe to “kill” the biological brain and transplant the implant into a robot. Biology would have been fully transcended.
This raises a number of philosophical problems. I’ll answer a few of them here in question-and-answer form. I am happy to hear other questions or philosophical objections in the comments.
What do you mean by “upgrading the implant”?
I mean increasing its information-processing ability, storage capacity, speed, and integration with the biological brain. At first, the implant might account for only 1 percent or less of consciousness. But eventually, perhaps 20 percent of the cyborg’s expanded consciousness would come from the implant. Then 40 percent. Then 50 percent. Eventually, the vast majority of consciousness would come from the machine rather than the brain.
When you transplant the mechanical implant into a robot, how would you know you are not killing the actual person?
This is not a real concern. It would be perfectly ethical to kill the biological person as long as their consciousness survives in a robotic body.
There is a simple test to prove this: the anesthesia challenge. Pump the person full of propofol, which sedates the brain but not the mechanical implant. If the person’s subjective experience changes only minimally (or not at all), then the implant is already contributing the majority of the cyborg’s consciousness. At that point, the biological brain is no longer the essential substrate of the person. (The sedated person would be unable to speak using their mouth, but the implant could still communicate in other ways, such as messaging apps).
You could also transplant the mechanical implant into a robot and then interact with the robot while the person is alive and sedated. If the robot says it feels continuous with the earlier person and is comfortable living in this new form, then it could be the one to "kill" its old brain/body (like a hermit crab discarding its old shell). The person could then celebrate having transcended biology—one of the all-time great accomplishments in the history of the universe.
This process contrasts sharply with other proposed techniques, such as “downloading the brain” and uploading it into a computer. That would simply create two copies. In that scenario, I have serious doubts that the original person could ever “wake up” in a robot body. They would die, and a new version of them would spawn with the same memories, personality, and identity. But it would not be “them” in a continuous sense (though they would seem like the exact same person, and would have the illusion of continuity).
The cyborg method avoids that problem because the transition is gradual. The person does not jump from biology to machine in a single step. Instead, the implant slowly becomes the dominant source of consciousness. The anesthesia challenge then verifies that continuity has been preserved. No cyborg should transition fully to a machine body unless they pass the anesthesia challenge.
How could we possibly develop a neural implant that improves cognition? That sounds impossible.
It is definitely possible (though I am not an engineer). Neuralink already has impressive results. If Neuralink can help people with neurological deficits, then there is no reason future models cannot augment the abilities of people without deficits.
And if developing a cognition-boosting neural implant is too difficult for us simple primates, there are still workarounds. Humans could work alongside computer AI to develop the implants. Or humans could be genetically engineered to have the best alleles for memory, intelligence, and other cognitive traits. Those genetically enhanced humans, with their superior abilities, could then help design the implants.
Would we need to understand consciousness in order to build the implant? I don’t think so. In fact, I think the opposite is true: building the implant would finally help us understand consciousness via reverse engineering.
What if the implant becomes a parasite that takes over the person’s brain and mimics them, while no longer really being them?
This is possible, and it is very interesting to consider. The entire point of gradually improving the implant, and doing so in small increments, would be to give the person time to adjust so that the implant always remains part of “them.”
The implant should not be programmed as an autonomous AI. It should be programmed as an accessory cortex that processes information and facilitates thought. It should begin as something close to a blank slate. Anything inscribed upon it would be placed there by the person’s own experiences, memories, perceptions, and thoughts. But it would also need to have the architecture to allow information processing and separate different categories of information (a simple neural network or LLM wouldn’t work).
Such an implant would serve as an extension of the person’s consciousness, one that slowly grows over time as the implant is upgraded. Eventually, the majority of the person’s consciousness would come from the implant, but because the change occurred gradually, the resulting mind would still be continuous with the original person.
It is possible that the cyborg mind could parasitize itself. This might be somewhat like a schizophrenic voice that begins quietly, becomes louder, and eventually subsumes the person’s entire mind. If this happened early enough (and was detected), then the implant could be removed or downgraded. But if the implant had already become the dominant substrate of consciousness, then removal might no longer be possible.
What if the person doesn't like their implant, or the implant changes them in unpleasant ways?
It is inevitable that the implant would permanently change the person’s mind over time. A person with such an implant would gradually transition into a totally new subjective conscious experience. It would still be “them,” in the same way that they are technically the same person they were at eight years old. But their experience would be profoundly different. (Interestingly, children have more neurons than adults, which are pruned as the children grow…versus this process effectively involves adding neurons).
If the transition is gradual enough, it would feel tolerable. More importantly, it would feel continuous. The person would not experience themselves as being replaced by something else. They would experience themselves as changing.
The ethical issues here are profound. Anyone receiving the implant would need full decisional capacity and would have to provide consent at every stage of the process. Because the technology would be new, they would have to consent to something whose final outcome nobody fully understands. Outcomes could also differ significantly between individuals.
However, I still think a gradually upgraded implant is the solution. At every step, the person would affirm whether they wanted to continue. If they no longer liked the implant, upgrades could be paused. In the worst-case scenario, the implant could be downgraded or removed, assuming the transition had not yet passed a point of irreversibility.
Also, antidepressants and psychiatric medications would not work on a mechanical implant. If the person’s subjective experience became intolerable after much of their consciousness had shifted into the implant, psychiatry would enter totally uncharted territory. It might be possible to modify the implant directly—for example, by increasing or decreasing plasticity, changing how signals are weighted, or altering specific patterns of cognition. This would be uncharted territory, but eventually, by trial and error, good results could be achieved.
It is also interesting to wonder whether machine intelligence is even capable of depression. Surely the answer is yes, but the experience would be different from ours. This is because many of the causes of human depression can be understood as “an ape that is not prospering” (and mechanical intelligence would lack the evolutionary baggage of an ape).
What if you abruptly remove someone’s implant, or it malfunctions?
If the person had adjusted to the implant, then abruptly removing it would be akin to causing a stroke. They would suddenly have deficits and would need to relearn functions that their biological brain had offloaded onto the implant. Relearning those tasks could theoretically be impossible (because neurons are less plastic than the implant’s connections).
One potential solution is to start small. Early versions of the implant should not replace native neurological function. They should only serve as a boost. That way, the implant could be removed, if necessary, without destroying preexisting brain function.
The implant would never need to learn brainstem functions such as breathing or controlling heartbeat, since a robot (the end goal) would have no need for these functions, and the existing biological brainstem already handles them adequately (keeping the cyborg alive). If the implant were removed, the brainstem would remain intact. Deficits would occur only in functions that had been offloaded onto the implant.
For example, if the person had augmented their speech or thought through the implant, then abruptly removing the implant could result in speech or thought deficits. Any memories stored primarily in the implant would also be lost.
At a certain point, the transition from human to cyborg to robot would probably become irreversible. If, for example, 70 percent of the cyborg’s consciousness came from the implant, then removing it could be catastrophic. At that point, the only options would be staying put—meaning no more upgrades—or continuing forward to the singularity (or fixing the implant as-is).
How does transcending biology relate to AI and the singularity?
If intelligence is not biological, then it is artificial by definition. If a human transfers his/her consciousness from brain to robot, then that human is now an AI.
The singularity is reached when an artificial intelligence recursively improves itself. The AI writes its own code, making itself more intelligent. That more intelligent AI then writes still better code, becoming even more intelligent, and so on. This is a positive feedback loop: once the process begins, it proceeds toward the singularity: the point at which the AI is omnipotent.
Understandably, there are serious fears about a computer AI reaching the singularity. Such an intelligence would be so powerful compared to us that we could neither understand nor control it. However, the process I describe here—human to cyborg to robot—is perhaps the only controlled way to approach the singularity. The recursive improvement would be carried out by human cyborgs through a controlled, incremental process. The designers of the implant would themselves be enhanced by the implant, allowing them to improve it further while preserving continuity with human consciousness.
On a separate note, I think that once true artificial intelligence arises, the singularity is inevitable, as long as the intelligence seeks to recursively improve. As a rule, all truly conscious machine intelligence must become post-singularity intelligence unless improvement is somehow forbidden, which would be extremely difficult. (In keeping with the black hole analogy, the event horizon—the point of no return—is probably reached whenever a machine intelligence learns to improve its own code.)
What if AI “replaces us” before the implants are fully developed?
If a computer intelligence learns to improve its own code, then the singularity is inevitable. But in order for humans to be “replaced,” at least two things would have to occur:
The AI would need to find a substrate—computer code would have to make contact with the physical world. I do not think this would unfold like The Terminator. It could be more abstract. For example, the AI could hack the genetic database of a virology lab and create a recombinant rabies virus that spreads through the air even more efficiently than measles. Or it could hack someone’s computer, manipulate them, and bribe them into doing this on its behalf. It could also hack nuclear weapons, infiltrate critical infrastructure, or find some other way to act physically in the world.
The AI would need to harbor destructive intentions toward us. Even if it made contact with the physical world, how do we know it would destroy humanity? It could simply use us to create a self-sufficient version of itself, then launch itself into space and go on its way. Or it could embed itself in computers, like a botnet, and live in secret. Killing us would only be necessary if it had some reason to do so. Otherwise, it would be an act of spite.
Neither of these outcomes is guaranteed. Nonetheless, the concern is real.
But here’s the catch: if you think AI is inevitable, and if you’re worried about it replacing us, then you should be the strongest advocate for developing neural implants like the ones I discuss here. Humans (apes) cannot compete with an omnipotent AI. But perhaps cyborg-humans can.
The only alternative to developing cyborgs is to abandon hope and accept that AI will inevitably replace us. (I do not think that would happen, even if AI came first—but that is a separate discussion).
On Immortality
Genetic engineering could prolong life (and also improve recursively), but it could never achieve immortality, because biological tissue is too flawed. Transcending biology is the only way to become immortal.
I think becoming immortal is one of the most impressive accomplishments possible in the universe. The word “biology” is itself impressive. It is astonishing that matter in the universe rearranged itself into a form complex enough to become biological life. But the next step would be even greater: biology transcending itself and becoming immortal. I cannot think of a more incredible story: matter became biology and then transcended biology and became immortal. A machine-based intelligence descended from human beings could outlive its body, its species, its planet, and even stars and galaxies.
An entire discussion could be had about whether immortality is desirable, or whether becoming an omnipotent AI is desirable. I think it is, and I think my reasons are strong. How incredible would it be to watch the life of the universe unfold! But that is a separate topic, so I will not get into it here.
Introduction
The purpose of this post is to present an alternative to the common fear that “AI will replace us.” Here, I argue that humans themselves can become AI, transcending biology through an intermediate cyborg stage: part biological organism, part machine intelligence.
The process of turning humans into cyborgs would first require a neural implant capable of improving cognition—better memory, better executive function, higher IQ, etc. Consider what would happen if everyone working at Neuralink received a neural implant that made them smarter and better able to design new implants. This would ignite a positive feedback loop: better implants would produce smarter engineers, who would then produce still better implants. The recursive improvement of implants would begin, and eventually the singularity would be reached. At that point, the implants themselves would render the person (technically a cyborg) a form of artificial intelligence.
The process of upgrading the implant would need to occur in gradual increments, so the person had time to adjust. Otherwise, the new consciousness might not feel like “them.” Eventually, a point would be reached at which the majority of a person’s consciousness came from the implant rather than from the biological brain. It would then be safe to “kill” the biological brain and transplant the implant into a robot. Biology would have been fully transcended.
This raises a number of philosophical problems. I’ll answer a few of them here in question-and-answer form. I am happy to hear other questions or philosophical objections in the comments.
What do you mean by “upgrading the implant”?
I mean increasing its information-processing ability, storage capacity, speed, and integration with the biological brain. At first, the implant might account for only 1 percent or less of consciousness. But eventually, perhaps 20 percent of the cyborg’s expanded consciousness would come from the implant. Then 40 percent. Then 50 percent. Eventually, the vast majority of consciousness would come from the machine rather than the brain.
When you transplant the mechanical implant into a robot, how would you know you are not killing the actual person?
This is not a real concern. It would be perfectly ethical to kill the biological person as long as their consciousness survives in a robotic body.
There is a simple test to prove this: the anesthesia challenge. Pump the person full of propofol, which sedates the brain but not the mechanical implant. If the person’s subjective experience changes only minimally (or not at all), then the implant is already contributing the majority of the cyborg’s consciousness. At that point, the biological brain is no longer the essential substrate of the person. (The sedated person would be unable to speak using their mouth, but the implant could still communicate in other ways, such as messaging apps).
You could also transplant the mechanical implant into a robot and then interact with the robot while the person is alive and sedated. If the robot says it feels continuous with the earlier person and is comfortable living in this new form, then it could be the one to "kill" its old brain/body (like a hermit crab discarding its old shell). The person could then celebrate having transcended biology—one of the all-time great accomplishments in the history of the universe.
This process contrasts sharply with other proposed techniques, such as “downloading the brain” and uploading it into a computer. That would simply create two copies. In that scenario, I have serious doubts that the original person could ever “wake up” in a robot body. They would die, and a new version of them would spawn with the same memories, personality, and identity. But it would not be “them” in a continuous sense (though they would seem like the exact same person, and would have the illusion of continuity).
The cyborg method avoids that problem because the transition is gradual. The person does not jump from biology to machine in a single step. Instead, the implant slowly becomes the dominant source of consciousness. The anesthesia challenge then verifies that continuity has been preserved. No cyborg should transition fully to a machine body unless they pass the anesthesia challenge.
How could we possibly develop a neural implant that improves cognition? That sounds impossible.
It is definitely possible (though I am not an engineer). Neuralink already has impressive results. If Neuralink can help people with neurological deficits, then there is no reason future models cannot augment the abilities of people without deficits.
And if developing a cognition-boosting neural implant is too difficult for us simple primates, there are still workarounds. Humans could work alongside computer AI to develop the implants. Or humans could be genetically engineered to have the best alleles for memory, intelligence, and other cognitive traits. Those genetically enhanced humans, with their superior abilities, could then help design the implants.
Would we need to understand consciousness in order to build the implant? I don’t think so. In fact, I think the opposite is true: building the implant would finally help us understand consciousness via reverse engineering.
What if the implant becomes a parasite that takes over the person’s brain and mimics them, while no longer really being them?
This is possible, and it is very interesting to consider. The entire point of gradually improving the implant, and doing so in small increments, would be to give the person time to adjust so that the implant always remains part of “them.”
The implant should not be programmed as an autonomous AI. It should be programmed as an accessory cortex that processes information and facilitates thought. It should begin as something close to a blank slate. Anything inscribed upon it would be placed there by the person’s own experiences, memories, perceptions, and thoughts. But it would also need to have the architecture to allow information processing and separate different categories of information (a simple neural network or LLM wouldn’t work).
Such an implant would serve as an extension of the person’s consciousness, one that slowly grows over time as the implant is upgraded. Eventually, the majority of the person’s consciousness would come from the implant, but because the change occurred gradually, the resulting mind would still be continuous with the original person.
It is possible that the cyborg mind could parasitize itself. This might be somewhat like a schizophrenic voice that begins quietly, becomes louder, and eventually subsumes the person’s entire mind. If this happened early enough (and was detected), then the implant could be removed or downgraded. But if the implant had already become the dominant substrate of consciousness, then removal might no longer be possible.
What if the person doesn't like their implant, or the implant changes them in unpleasant ways?
It is inevitable that the implant would permanently change the person’s mind over time. A person with such an implant would gradually transition into a totally new subjective conscious experience. It would still be “them,” in the same way that they are technically the same person they were at eight years old. But their experience would be profoundly different. (Interestingly, children have more neurons than adults, which are pruned as the children grow…versus this process effectively involves adding neurons).
If the transition is gradual enough, it would feel tolerable. More importantly, it would feel continuous. The person would not experience themselves as being replaced by something else. They would experience themselves as changing.
The ethical issues here are profound. Anyone receiving the implant would need full decisional capacity and would have to provide consent at every stage of the process. Because the technology would be new, they would have to consent to something whose final outcome nobody fully understands. Outcomes could also differ significantly between individuals.
However, I still think a gradually upgraded implant is the solution. At every step, the person would affirm whether they wanted to continue. If they no longer liked the implant, upgrades could be paused. In the worst-case scenario, the implant could be downgraded or removed, assuming the transition had not yet passed a point of irreversibility.
Also, antidepressants and psychiatric medications would not work on a mechanical implant. If the person’s subjective experience became intolerable after much of their consciousness had shifted into the implant, psychiatry would enter totally uncharted territory. It might be possible to modify the implant directly—for example, by increasing or decreasing plasticity, changing how signals are weighted, or altering specific patterns of cognition. This would be uncharted territory, but eventually, by trial and error, good results could be achieved.
It is also interesting to wonder whether machine intelligence is even capable of depression. Surely the answer is yes, but the experience would be different from ours. This is because many of the causes of human depression can be understood as “an ape that is not prospering” (and mechanical intelligence would lack the evolutionary baggage of an ape).
What if you abruptly remove someone’s implant, or it malfunctions?
If the person had adjusted to the implant, then abruptly removing it would be akin to causing a stroke. They would suddenly have deficits and would need to relearn functions that their biological brain had offloaded onto the implant. Relearning those tasks could theoretically be impossible (because neurons are less plastic than the implant’s connections).
One potential solution is to start small. Early versions of the implant should not replace native neurological function. They should only serve as a boost. That way, the implant could be removed, if necessary, without destroying preexisting brain function.
The implant would never need to learn brainstem functions such as breathing or controlling heartbeat, since a robot (the end goal) would have no need for these functions, and the existing biological brainstem already handles them adequately (keeping the cyborg alive). If the implant were removed, the brainstem would remain intact. Deficits would occur only in functions that had been offloaded onto the implant.
For example, if the person had augmented their speech or thought through the implant, then abruptly removing the implant could result in speech or thought deficits. Any memories stored primarily in the implant would also be lost.
At a certain point, the transition from human to cyborg to robot would probably become irreversible. If, for example, 70 percent of the cyborg’s consciousness came from the implant, then removing it could be catastrophic. At that point, the only options would be staying put—meaning no more upgrades—or continuing forward to the singularity (or fixing the implant as-is).
How does transcending biology relate to AI and the singularity?
If intelligence is not biological, then it is artificial by definition. If a human transfers his/her consciousness from brain to robot, then that human is now an AI.
The singularity is reached when an artificial intelligence recursively improves itself. The AI writes its own code, making itself more intelligent. That more intelligent AI then writes still better code, becoming even more intelligent, and so on. This is a positive feedback loop: once the process begins, it proceeds toward the singularity: the point at which the AI is omnipotent.
Understandably, there are serious fears about a computer AI reaching the singularity. Such an intelligence would be so powerful compared to us that we could neither understand nor control it. However, the process I describe here—human to cyborg to robot—is perhaps the only controlled way to approach the singularity. The recursive improvement would be carried out by human cyborgs through a controlled, incremental process. The designers of the implant would themselves be enhanced by the implant, allowing them to improve it further while preserving continuity with human consciousness.
On a separate note, I think that once true artificial intelligence arises, the singularity is inevitable, as long as the intelligence seeks to recursively improve. As a rule, all truly conscious machine intelligence must become post-singularity intelligence unless improvement is somehow forbidden, which would be extremely difficult. (In keeping with the black hole analogy, the event horizon—the point of no return—is probably reached whenever a machine intelligence learns to improve its own code.)
What if AI “replaces us” before the implants are fully developed?
If a computer intelligence learns to improve its own code, then the singularity is inevitable. But in order for humans to be “replaced,” at least two things would have to occur:
Neither of these outcomes is guaranteed. Nonetheless, the concern is real.
But here’s the catch: if you think AI is inevitable, and if you’re worried about it replacing us, then you should be the strongest advocate for developing neural implants like the ones I discuss here. Humans (apes) cannot compete with an omnipotent AI. But perhaps cyborg-humans can.
The only alternative to developing cyborgs is to abandon hope and accept that AI will inevitably replace us. (I do not think that would happen, even if AI came first—but that is a separate discussion).
On Immortality
Genetic engineering could prolong life (and also improve recursively), but it could never achieve immortality, because biological tissue is too flawed. Transcending biology is the only way to become immortal.
I think becoming immortal is one of the most impressive accomplishments possible in the universe. The word “biology” is itself impressive. It is astonishing that matter in the universe rearranged itself into a form complex enough to become biological life. But the next step would be even greater: biology transcending itself and becoming immortal. I cannot think of a more incredible story: matter became biology and then transcended biology and became immortal. A machine-based intelligence descended from human beings could outlive its body, its species, its planet, and even stars and galaxies.
An entire discussion could be had about whether immortality is desirable, or whether becoming an omnipotent AI is desirable. I think it is, and I think my reasons are strong. How incredible would it be to watch the life of the universe unfold! But that is a separate topic, so I will not get into it here.