> Calculators are deterministic and necessarily correct. GPT is probabilistic and not necessarily correct.
This isn't really relevant to the issue.
We have a choice.
(a) Technology is subject to the will of each user. If they want to make it do something bad, that's what it does.
(b) Technology is subject to the will of some organization. They try to control what the technology will do.
The first one is clearly better for multiple reasons.
We already have systems (e.g. courts) for punishing people who do sufficiently bad things, which mitigates the harms of the first one.
The alleged good from the second one often isn't realized, because the really bad people (e.g. adversarial foreign governments, major criminal organizations) aren't going to be subject to it or will hack the safeguards.
The imposition of controls comes with a level of mass surveillance that isn't compatible with a free society.
And putting anyone in charge of what everyone else can do or see is a dystopia because there is no one you can trust with that kind of power, for reasons of both incompetence and avarice.
Technology should be designed to distribute power, not concentrate it.