[zwaps]

My worry about AI isn't Skynet and Terminators.

It's that Data Scientists don't realize most of their models still simply fit a conditional expectation - and given all their power, it's going to be us, not Amazon or Google, who has to adapt to the distribution as not to repeatedly get hit as some sort of outlier.

It's a dystopia where AI works because of us trying to conform to it, because otherwise we are out of luck. At some point, we self select into Amazon-humans, Google-humans etc.

[317070]

> It's a dystopia where AI works because of us trying to conform to it, because otherwise we are out of luck

But all technology is like that. Cars work, because we conformed our environment to it. Email works, because we conform to checking our mailboxes regularly. Telephones work, because we pick them up when they ring.

If you are interested in knowing more, the whole work of the french philosopher Jacques Ellul is based around this idea.

AI is not going to be any different in this respect. It is going to be us meeting it halfway. And that is going to be dystopic in the same way as it has been with other types of technology.

[alistairSH]

Wow, not sure why you're getting all the down-votes. I don't necessarily agree with you, but it's an interesting perspective regardless.

You also touch on the Amish approach to new technology. Each new thing is assessed for impact on life/tradition/etc and adoption is limited accordingly. For example, phones might be considered useful enough to have one per town, but distracting enough to be kept out of individual homes. Or, a computer might be allowed in the back-office of a wood shop for purposes for managing online sales, but nowhere else in the community.

[andrewzah]

The difference is we have a choice in those matters. I.e. my telephone does not work in that sense, because I will not pick up for an unknown number in most cases. A communicable human is on the other end.

With "AI" (machine learning / neural nets), we're letting opaque boxes arbitrarily control and ban accounts at will. It's like forcing people to use cars that randomly decide that it won't drive that day for no good reason.

"AI is not going to be any different in this respect"

The other examples all include accountable humans, "AI" here does not. We're already seeing issues of machine learning applying sexist biases for example because of datasets and poor training. What are you going to do, hold google, or any other faceless corporation, accountable for its bad translations? Good luck.

It's only going to get worse in the future as other stratified aspects of society get cemented by "AI".

[317070]

So, I think I explained myself badly. I fully agree on the problems of AI you point out. I agree with you on the reduction of freedom AI will bring. But then again, every relevant piece of technology took away such a piece of freedom.

I think one might underestimate the freedom that have been let go when other technologies came in, even though those technologies feel established already. You can choose not to pick up on the unknown number when the phone rings, but the moment of connection with a real human being you might have had at that point was interrupted by the phone ringing.

Society expects us to be reachable at any time, but that came with trading in the freedom to not be arbitrarily disrupted by default.

The point I am trying to make is this: all technology requires and has required humans to adapt to it, and consequently handing in a bit of freedom. I want to point out that AI is no different in the fact that it does so, even though like every other piece of technology it will do so in a unique way.

Is that dystopian? Perhaps. Many philosophers on technology have thought so, and many have disagreed.

> The other examples all include accountable humans

But when those faceless corporations were being sexist because of bureaucratic reasons in the sixties, how were they being held accountable? Did anything really change?

Widely adopted technology is something that lies outside of the humans that make it up, like an ant colony is different from the ants it is made off. You cannot hold any individual ant accountable to the way the hive is organised. Technology is an entity that lies outside of those ants, and that has its own desires and goals.

[technofiend]

>At some point, we self select into Amazon-humans, Google-humans etc.

That's one worst case scenario. I've been thinking once humans start deselecting Google (no more gmail, youtube red, etc) then there's more room to also start blocking google's ad network. I know Google's dark pattern to force compliance is to pretend they don't know you and put you on a captcha treadmill, but that too will crumble if their adsense network yields less and less value as users and website operators move on from being plugged into the googleverse.

At least today you can operate outside the googleverse but you're punished for it. We should be holding on to that and trying to expand it rather than end up in the future you described where we can do nothing but comply.