I haven't seen any AI whose information I would trust. They're all being fed garbage as inputs and, increasingly, some of that garbage is slop from other AIs.
If you extrapolate that out, there's no way to ensure AI won't be outputting crap at least some of the time.
You would have to build an AI from scratch with 100% human-curated inputs (try that without any bias), which would be both impractical and time-consuming, and then you'd have to do the same with any ongoing inputs to keep it on track until it was mature enough to discern for itself.
I haven't seen any AI whose information I would trust. They're all being fed garbage as inputs and, increasingly, some of that garbage is slop from other AIs.
If you extrapolate that out, there's no way to ensure AI won't be outputting crap at least some of the time.
You would have to build an AI from scratch with 100% human-curated inputs (try that without any bias), which would be both impractical and time-consuming, and then you'd have to do the same with any ongoing inputs to keep it on track until it was mature enough to discern for itself.
Wake me up when that happens.
Some people have conversations with Groken AI.