" Right now, we’re much smarter than AI and fully in control of the thing that is in front of us."
So simple: Do not give up control.
Example: Atom rockets should not be startet by an AI itself.
Naturally they should not be startet at all. To prevent their start we need several human beeings in the command chain.
Also we should be aware of our schizophrenic attitude. We want the AI to be creative but at the same time we want to maintain controll. Results should be universal but we censor.
It all boils down to OUR responsibility, but I fear that greed and power mongering is to strong, as also indicated by Jurgen Gravestein.
Food for thought. If A.I. is helping to align other AI where does that logically lead? I mean doesn't RLAIF (and its evolution) replace RLHF inevitably.
" Right now, we’re much smarter than AI and fully in control of the thing that is in front of us."
So simple: Do not give up control.
Example: Atom rockets should not be startet by an AI itself.
Naturally they should not be startet at all. To prevent their start we need several human beeings in the command chain.
Also we should be aware of our schizophrenic attitude. We want the AI to be creative but at the same time we want to maintain controll. Results should be universal but we censor.
It all boils down to OUR responsibility, but I fear that greed and power mongering is to strong, as also indicated by Jurgen Gravestein.
Alignment - to what? That's a real rub you're pointing out. It makes me shudder to think of politics playing out here but I suppose that's inevitable.
Food for thought. If A.I. is helping to align other AI where does that logically lead? I mean doesn't RLAIF (and its evolution) replace RLHF inevitably.
I think of the alignment question as a debate about our own values. Which ones should drive the development of current tech?