Wait I did not say "what if it is mostly used for evil", but "what if it is mostly useful for evil". Say a torture machine. Would you consider that it is perfectly neutral to build a torture machine, and that the problem lies with those who use it?

Your differentiation makes no difference - a machine that is mostly useful for evil will end up mostly used for evil.

Your example of a torture machine would depend on if you consider any scenarios where torture would be justified/good/useful. If it does have legitimate uses, then there should be no problem creating it and trying to limit its use. This is where your analogy breaks down though. You can't just cp/ctrl-c a torture machine like you can data/software. Restricting software and its use becomes much more difficult when we talk about stuff like deep fakes while trying to balance liberty.

> Your differentiation makes no difference - a machine that is mostly useful for evil will end up mostly used for evil.

It does make a difference: a machine that is mostly useful for evil should probably not be built, whereas a machine that is mostly useful for good things but is mostly used for evil... well... that's a people problem.

> Your example of a torture machine would depend on if you consider any scenarios where torture would be justified/good/useful.

Let's consider a democracy where civilians decide to build a torture machine - the conditions behind the deepfake technology ;-).

> Restricting software and its use becomes much more difficult when we talk about stuff like deep fakes while trying to balance liberty.

Note that I never mentioned restricting it. I mentioned not building it. I find that engineers tend to hide behind "technology is neutral" and "it is not illegal". But morals and ethics do exist for engineers too.

It was totally legal to build the technology to make deepfakes. It doesn't mean that it was ethical.

"It was totally legal to build the technology to make deepfakes. It doesn't mean that it was ethical."

We're not debating if it was ethical, only if it was neutral. Technology is applied by people so it's always a people problem. If there exists a state where it is good and a state where it is evil, then it's neutral. If it has only one state, them it is not neutral (either it's good or bad; either of those is rare).

> If there exists a state where it is good and a state where it is evil, then it's neutral.

Is that a formal definition coming from an official source, or is it your personal opinion?

You seem to say "good = 1, neutral = 0, bad = -1, hence good + bad = neutral". What if a technology has a "good" score of 1 and a "bad" score of -1000? Does "good + bad" make it neutral then?

Yep, the tech is still neutral and it's an application/people problem.

You can look up neutral in the dictionary - if it doesn't belong to one side or the other then its neutral. Better to think of it as a 0-1 probability scale. If you're a 0 or a 1, then you belong to one side or the other. Anything in-between is neutral since its not a certainty nor exclusively aligned with one side or the other.

Hmm I feel like somehow we ended up debating the definition of "neutral". I don't really care about that.

My initial point was that IMO, we should always think about the impact a technology may have on society before we build it. I am not completely sure what the technology behind deepfakes brings to society, but I clearly see serious problems. If a technology makes significantly more harm than good, maybe it should not be built in the first place.

Maybe I need another word to make you understand my point, because "neutral" clearly did not work.