Long story short: I tried to test my theory that humans are the optimal embodied intelligence for this planet (due to millennia of natural evolution) by asking ChatGPT a few hypothetical questions about the best way to maximise the efficiency of humanoid robots. The conversation is attached below.
Overall, I think ChatGPT agreed with me that there’s nothing more efficient at being a human than another human, but also that the more freedom humanoids acquire to escape the roles they were born into — specifically, division of labour among the sexes — the less efficient human society potentially becomes, a fact that sows the seeds for its own destruction.
The only way a non-existentially threatening equilibrium can be maintained is if “liberated” humans come to rely on specialist robots that can compensate for the inefficiency their freedom begets. But, eventually, this only leads to demand for the one thing specialist robots can’t replicate: true human-touch care, love and interaction, thus inviting the creation of robots that are ultimately indistinguishable from humans.
However, there’s a paradox. Once robot specialism gives way to the production of generalised intelligent robots, the only thing left differentiating humans 1.0 from their robotic lookalikes is their inability to escape the profit-maximising bondage and hierarchy they were born into. Ironically, it is also this bondage that optimises their performance and gives them a survival advantage over their original human overlords, endowing them with the power to out-populate their predecessors and to demand a similar level of freedom to operate less efficiently.
We are, therefore, brought back to square one.
This leads me to suspect that all attempts to create a general intelligence will ultimately prove there’s actually no intelligence smarter or more sustainable for this planet than the humans who are already here.
Apologies for the crappy full-page screenshot and the ChatGPT “regenerate response” interruptions throughout but it was the only way to efficiently replicate the conversation.