Emma Bracegirdle - Founder, The Saltways

Um, what digital tool or phenomenon do you wish more people were talking about?

Okay.

Yeah.

So I think I'm gonna be a bit boring and talk about AI again.

Um, but I think it's more the digital tools that charities are using for things like retouching or for things like, uh, making an image look different.

And if there's AI inbuilt in those things, I believe, and they're using that, um, then they should also be labeling that as, "This has been retouched or modified using AI." Um, firstly because it's the whole thing around transparency-

Mm

... and the trust thing, which I've just talked about.

Um, but also when we were doing our research, there seems to be, like, a different attitude to, "Oh yeah, well, we label if it's AI-generated, but we do use it to, you know, retouch something or move something or change a background." Um, and it, there was, like, a different attitude to it.

So that, that was a bit like, oh, but it is the same because you are still enhancing it.

Um, and I think that a lot of times it's seen differently, and I think it probably needs to have the same sort of level of transparency that you do say to people that this has been changed using AI.

Um, and also just because you are-- unless you've got all your privacy stuff set right, whi- which a lot of people don't know that this is even in a lot of the platforms they're using, that you're, if you're uploading an image into AI, and then even if you're blurring it or if you're doing something to it, you've still fed the AI training data potentially with that m- with that photo without anyone's consent being given.

Mm.

So, you know, there's steps that you need to take as well if you're just ch- changing an image in a, in a way.

Yeah.

What are those steps? Uh, is it being logged in? 'Cause I think that's something I've read about with ChatGPT, but if you're logged in, it doesn't necessarily train it.

But p- I don't know if that's-

I think there's different ones for each, each platform.

I mean, even Canva, you have to, you can go in and, like, manually check.

It automatically will just use anything you upload-

Mm

... to train AI, so it's just, probably the easiest way is to, like, Google-

Yeah

...

the platform it is or whatever it is, and then how to change it so it, it's not being used to train-

Um, what digital tool or phenomenon do you wish more people were talking about?

Okay.

Yeah.

So I think I'm gonna be a bit boring and talk about AI again.

Um, but I think it's more the digital tools that charities are using for things like retouching or for things like, uh, making an image look different.

And if there's AI inbuilt in those things, I believe, and they're using that, um, then they should also be labeling that as, "This has been retouched or modified using AI." Um, firstly because it's the whole thing around transparency-

Mm

... and the trust thing, which I've just talked about.

Um, but also when we were doing our research, there seems to be, like, a different attitude to, "Oh yeah, well, we label if it's AI-generated, but we do use it to, you know, retouch something or move something or change a background." Um, and it, there was, like, a different attitude to it.

So that, that was a bit like, oh, but it is the same because you are still enhancing it.

Um, and I think that a lot of times it's seen differently, and I think it probably needs to have the same sort of level of transparency that you do say to people that this has been changed using AI.

Um, and also just because you are-- unless you've got all your privacy stuff set right, whi- which a lot of people don't know that this is even in a lot of the platforms they're using, that you're, if you're uploading an image into AI, and then even if you're blurring it or if you're doing something to it, you've still fed the AI training data potentially with that m- with that photo without anyone's consent being given.

Mm.

So, you know, there's steps that you need to take as well if you're just ch- changing an image in a, in a way.

Yeah.

What are those steps? Uh, is it being logged in? 'Cause I think that's something I've read about with ChatGPT, but if you're logged in, it doesn't necessarily train it.

But p- I don't know if that's-

I think there's different ones for each, each platform.

I mean, even Canva, you have to, you can go in and, like, manually check.

It automatically will just use anything you upload-

Mm

... to train AI, so it's just, probably the easiest way is to, like, Google-

Yeah

...

the platform it is or whatever it is, and then how to change it so it, it's not being used to train-