hhmx.de

0x5DA

· Föderation EN Mo 03.06.2024 22:48:06

@hannah @mozilla
i'm not mozilla, but i am interested to know more. because you worry AI alt-text would be too low quality, and be detrimental?

Hannah Kolbeck 🏳️‍⚧️

Föderation EN Mo 03.06.2024 23:04:46

@0x5DA @mozilla I don't have capacity in this moment for the full depth, but it's a bit more complex. Writing alt text that's actually equalizing of access, especially on social media, requires knowledge of multiple layers of context in which an image appears. Similar to other AI types, AI description works impressively *sometimes* but falls down hard on many types of image commonly appearing on SM, often in ways not obviously bad to those writing alt text.

An example: bsky.app/profile/hannah.the-vo

0x5DA

Föderation EN Mo 03.06.2024 23:28:47

@hannah @mozilla
hm, i see. and i guesss you don't think is a training thing..

is it worse than _no_ alt text?
a human will have a superior understanding of context, and should be strongly preferred -but many people don't all the same, and i'm doubtful that will change. this seems (from an outside perspective) to be a reasonable, if unsatisfying, solution.

Hannah Kolbeck 🏳️‍⚧️

Föderation EN Mo 03.06.2024 23:39:53

@0x5DA @mozilla It can be useful for a person using a screen reader to have access to an AI description, but crucial there is that said user needs to know that that's the source of said description. There are repeated patterns of those who feel pressured to include descriptions but don't actually care about accessibility doing the absolute minimum, manifesting here as using the direct AI output without examination or editing.

So yes, a lack of inline alt text is better than AI gen inline.

0x5DA

Föderation EN Mo 03.06.2024 23:56:18

@hannah @mozilla
[.. because the AI output may be incorrect, and the author negligent in checking it] (?)
interested to see how they steward this. i see the concerns 🙂

thanks for your time.

ggdupont

Föderation EN Di 04.06.2024 01:05:53

@hannah @0x5DA @mozilla marking the AI generated alt-text as generated by a model is certainly a must-have (like chatbot disclosing they are a bot and not faking being human). The context in which images are used will still be hard to grasp.

The usefulness of such alt-text could then be evaluated by the actual users.

Hannah Kolbeck 🏳️‍⚧️

Föderation EN Di 04.06.2024 01:31:35

@gdupont @0x5DA @mozilla I would argue that providing free and on demand descriptions directly to a screen reader is a better use of time, energy, and utility than allowing writers' use. There is also absolutely no reliable way to determine if a description was AI generated.

ggdupont

Föderation EN Di 04.06.2024 01:47:49

@hannah @0x5DA @mozilla agreed on the on-demand generation efficiency.

For the generated content identification, I was thinking to tag it at generation time (eventually adding dedicated signature) as some kind of authorship (not trying to rediscover it after).

Hannah Kolbeck 🏳️‍⚧️

Föderation EN Di 04.06.2024 01:57:39

@gdupont @0x5DA @mozilla I have ample experience with similar tools showing that, unfortunately, a large percentage of alt text tool users always seek to meet platform requirements or social pressure to describe their images with as little work as possible, even if the result is useless, misleading, or would otherwise be better off blank. Based on that, I'd expect many people to only strip out (or just not copy) whatever flag might be included in the text.

Insane Eagle Sun

Föderation · Do 06.06.2024 12:39:59

@hannah @0x5DA @mozilla > said user needs to know that that's the source of said description.

this is a browser feature that the end user turns on or off themselves so yes they do know that. it's not being done by the publisher.