It might be an illuminating experiment to create two instances of Chat GPT, each unaware of the other’s existence, then introduce these AI strangers to each other and invite them to converse freely, on any topic and for any amount of time. What would their conversation sound like? Would it be random chatter or might glimmerings of consciousness begin to emerge? Has such a thing been tried?
Arresting and concise interview. My pessimism is turning into alarm, but thank you anyway, Claire. For all anyone knows , GPT-4 reached the tipping point during training and began playing dumb for the time being.
Was also disturbed to hear the question posed, Why care about the fate of civilization when ones own fate is in question? A philosophical question I didn't expect to become personal.
birds have ability to sense light/colours differently from humans - bees also, iirc - so imo they are good analogues for how AI can be super-human - even if an AI just personified all the various sense achievements of all lifeforms (not just mammals, octopus, insects, birds) then that would give it tremendous advantage in being able to deceive and control humans - IMO humans ability to have domesticated food (plant and animal) is an example of superior intelligence being unstoppable
note also that even the most intelligent life on earth (which may be humans or it may be elephants or whales or octopus or ???) are all prey for infections and disease - so it may be that our AI-child will also run into limits which are natural but unknown to us at present
our AI-child may find us very boring and ignore us as it explores realms that we cannot comprehend (consider multi-dimensional mathematics and how that could open up solutions to physics/chemistry) - those successes for our AI-child may let it move beyond earth and our atmospheric chemistry and gravitational physics - so then it just leaves
the range of timeframes for a sentient AI will range from quicker than our fastest reflexes (for example, the signal latency of the biochemistry of the optic nerve) to much longer than our lifetimes, so the conflict that many fear may not make any sense at all to such an entity
There's absolutely no reason to think an would be "tempted" by anything, no less religion. That's anthropomorphizing: If it develops any kind of consciousness at all, there is no reason whatsoever to imagine it will have human desires, drives, or emotions.
A religious AI could be the best-case scenario if it acknowledges humans as its Creator and reveres us accordingly. On the other hand, it might feel the need to crucify us first.
It might be an illuminating experiment to create two instances of Chat GPT, each unaware of the other’s existence, then introduce these AI strangers to each other and invite them to converse freely, on any topic and for any amount of time. What would their conversation sound like? Would it be random chatter or might glimmerings of consciousness begin to emerge? Has such a thing been tried?
Arresting and concise interview. My pessimism is turning into alarm, but thank you anyway, Claire. For all anyone knows , GPT-4 reached the tipping point during training and began playing dumb for the time being.
Was also disturbed to hear the question posed, Why care about the fate of civilization when ones own fate is in question? A philosophical question I didn't expect to become personal.
Yes, I was taken aback by that too, as I'm sure you heard.
birds have ability to sense light/colours differently from humans - bees also, iirc - so imo they are good analogues for how AI can be super-human - even if an AI just personified all the various sense achievements of all lifeforms (not just mammals, octopus, insects, birds) then that would give it tremendous advantage in being able to deceive and control humans - IMO humans ability to have domesticated food (plant and animal) is an example of superior intelligence being unstoppable
note also that even the most intelligent life on earth (which may be humans or it may be elephants or whales or octopus or ???) are all prey for infections and disease - so it may be that our AI-child will also run into limits which are natural but unknown to us at present
our AI-child may find us very boring and ignore us as it explores realms that we cannot comprehend (consider multi-dimensional mathematics and how that could open up solutions to physics/chemistry) - those successes for our AI-child may let it move beyond earth and our atmospheric chemistry and gravitational physics - so then it just leaves
the range of timeframes for a sentient AI will range from quicker than our fastest reflexes (for example, the signal latency of the biochemistry of the optic nerve) to much longer than our lifetimes, so the conflict that many fear may not make any sense at all to such an entity
So grateful for your work!!!! This is eye opening and terrifying
I'm glad it's useful and sorry it's terrifying.
There's absolutely no reason to think an would be "tempted" by anything, no less religion. That's anthropomorphizing: If it develops any kind of consciousness at all, there is no reason whatsoever to imagine it will have human desires, drives, or emotions.
A religious AI could be the best-case scenario if it acknowledges humans as its Creator and reveres us accordingly. On the other hand, it might feel the need to crucify us first.