Re: AI.
I'm on the fence when it comes to AI when it comes to accessing the internet to learn. And here's where the line is for me:
-- if it is accessing paid sites, or places where you're expected to log in and/or spend money to access places, that shouldn't be on the list of things it can learn from.
-- if it is something open to the public, that anyone with eyeballs has access to... something you can read without a subscription, something you can watch (like YouTube), or similar, then I more or less consider it fair game.
Part of the 'right to be forgotten' thing in the UK came from some guy making a program which pieced together a person's identity from information available to the public. And what he was doing wasn't illegal - anyone with the time and some effort could do what his program did.
The issue: it did it TOO QUICKLY. It completely cut the time and effort that a normal human would put into searching for information on a person and putting it all together.
And the courts ruled THAT to be the problem... if a person did it on their own time and worked on it, sure, fine. The moment it became easy, it was an issue, because just anyone could do it then.
And, a part of me is seeing it reflected in the AI situation. An AI can use hundreds of thousands of bits of writing, artwork, media, whatever, and draw upon it all in an instant to do what it does. Of course, a human being can do the same thing ... but that takes time and effort, and most people don't have the skills, either.
The AI 'learns' much as we do - bits and pieces here and there, putting it together, then figuring out our own thing. Not everyone takes the time to do it well - we're not all artists. But the AI can take what it's learned, and put it to use in a flash.
It's doing what we do, but faster, and without the need for the proper skills. It's the same 'sin'.
But hasn't this always been the sin of machines? Jobs which used to require humans are being done more and more by machines, which can do it more efficiently, without requiring years of training to get to where it is. How are we for turning the clock back on that?
I get it. People are staking their livelihood on this. But that's always been the case since the first person dropped a machine into place to replace a human being. I don't think we're ever going to go back.
And I'm worried about the litigation. Yes, there should be restrictions on where an AI can go to learn - but for me that restriction is the registration / paywall. If it's out in the open? It's out in the open. If it's something a normal person (without registration / subscription) can access, it should have access.
Is it fair?
I don't know. The guy who wrote that program to get all the information that's publically accessible on anyone... was it fair that he could do in an instant what it would normally take days to do?
Was it fair to shut that down? When, again, anyone could do it?
And that's why I'm on the fence.
Emotion vs Reason. And I always have trouble with that, because a part of me goes 'yeah, I feel it, I can see what's at stake' and a part of me goes 'yeah, but rationally...'
And that side doesn't care about 'fair' in the emotional sense. It cares about ... consistency. If a person can do it, a machine should be allowed to do it, too.
"But it isn't creating anything"
No, not yet it isn't. Some day, it will. But what it's doing is adapting a bunch of information it's gained, sorting through it, and choosing the 'best' outcomes it can reason. Anyone who's spent years learning off another person's style is doing similar things. They're adding their own touches over time - their own style to it - but they learned on the backs of those before them.
AI is moving to that point, I think, and I think people don't LIKE that. But it's going to happen. A part of me feels this is the usual 'machines are taking our jobs', and the significant backlash is because it's in ART, which is the domain of 'humanity'. But if we're going to get fully cognizant AI at some point... Art isn't going to be off the table, now is it?
And what are those AI to learn from?
No comments:
Post a Comment