Is it possible that the fact we are still alive means that there is a core problem to the idea of existential risk from AI?
There are people who think that we already have AGI, and this number has only grown with the recent Bing situation. Maybe we have already passed the threshold for RSI, maybe we passed it years ago.
Is there something to the idea that you can slightly decrease your pdoom for every day we are still alive?
It seems possible to me that AI will just get better and better and we’ll just continue to raise the bar for when it is going to kill us, not realising that we have already passed that point and everything is fine for some reason.
I’m not saying I think this is the case, but I do consider it a possibility.
Current AI systems aren’t capable of doing large-scale software development in non-AI domains. Eg there is no AI-written operating system, compiler, or database (even though minimal versions of these are often done by individual students as single semester capstone projects). I don’t think we can infer much of anything from the fact that recursive self improvement hasn’t happened yet until after that threshold is crossed.
Is it possible that the fact we are still alive means that there is a core problem to the idea of existential risk from AI?
There are people who think that we already have AGI, and this number has only grown with the recent Bing situation. Maybe we have already passed the threshold for RSI, maybe we passed it years ago.
Is there something to the idea that you can slightly decrease your pdoom for every day we are still alive?
It seems possible to me that AI will just get better and better and we’ll just continue to raise the bar for when it is going to kill us, not realising that we have already passed that point and everything is fine for some reason.
I’m not saying I think this is the case, but I do consider it a possibility.
Current AI systems aren’t capable of doing large-scale software development in non-AI domains. Eg there is no AI-written operating system, compiler, or database (even though minimal versions of these are often done by individual students as single semester capstone projects). I don’t think we can infer much of anything from the fact that recursive self improvement hasn’t happened yet until after that threshold is crossed.