@eshep

I've not yet seen any posts, attacking or otherwise, taking this stance

Not yet.

Right now, this feature is only available in a development version of Mastodon and only on mastodon.social. Next will be the daredevil servers that run development code to be ahead of release in features.

But eventually, a new stable release will be rolled out with this feature. More and more servers will be upgraded to this new version with this feature. Once typical activist servers like beige.party introduce it, my prediction will come true.

I have family who have varying levels of blindness who consume media from all over the internet. All of them have been quite thrilled about the state of these new robot assistants we all now have such easy access to.

Specifically assigned "alt-text" is never necessary.

People who aren't sighted are usually happy about image-describing AI because it's better than nothing, and without image-describing AI, they've got literally nothing.

In fact, however, image-describing AI is just barely better than nothing, if at all. It describes stuff that doesn't matter. It doesn't describe stuff that does matter. Most importantly, it's unreliable and inaccurate. It hallucinates and describes stuff wrongly. However, people who aren't sufficiently sighted don't know. They can't verify whether what the AI says is true because they can't see the image well enough to be able to compare it with the description.

100% hand-written alt-text is always more reliable, more accurate and more fit for the context than AI-generated alt-text will ever be.

I know from first-hand experience. I've tasked an image-describing AI with describing an image which I've described manually with no AI help first, and then I've analysed the AI description and compared it with both the actual image and my descriptions. I've even done that twice. The results were abysmal.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Fediverse #Mastodon

So there's a new attack vector for Mastodon users against non-Mastodon users in the making.

Mastodon is rolling out alt-texts for what we on Hubzilla call the profile photo and the cover photo. It won't be long until it becomes mandatory in Mastodon's culture to have alt-texts for both these pictures. Regardless of how new this feature is and how many other features that were introduced between March, 2022 and now have not made it into Mastodon's culture and Mastodon's unwritten rules, by the way.

At this point, Mastodon's alt-text activists will start attacking anyone and everyone whom they catch without alt-texts for these two pictures. At least they'll lecture them about the importance of alt-texts. Or they'll block them outright.

What they won't realise:

• Not everyone in the Fediverse is on Mastodon. Just because you see something on your Mastodon Web UI or in your Mastodon app, doesn't mean it's on or from Mastodon itself.
• In fact, that particular user might not be on Mastodon.
• Just because Mastodon rolls out a new feature, doesn't mean everything in the Fediverse rolls out the same feature at the same time. This means that there's a whole lot of Fediverse server software that does not offer alt-texts for profile pictures, and that probably won't offer them for quite a while or ever.
• Also, Mastodon's culture is not and will never be the culture of the whole Fediverse. Sorry, Mastodon fundamentalists, but some software has been here before Mastodon, its culture is older than Mastodon itself, and it's technologically incompatible with Mastodon's unwritten rules.

Here on Hubzilla, where I'm posting from right now (in case you really thought I'm on Mastodon just like you), there is no such thing as a dedicated field for alt-texts anywhere. Even if you want to post pictures, there is no alt-text entry mask, and there is no alt-text database field for the images.

Instead, both images and alt-texts are handled like on a blog: You embed the image somewhere in the post text using markup code. In a sense, you program the image into the post. And if you want the image to have alt-text, you have to program the alt-text into the image-embedding code. If you're afraid of coding, Hubzilla is not for you.

But if there's no entry mask and no database field for alt-texts in posts, there won't be either for the profile images either. That is, unless Hubzilla adopts the alt-text field from (streams) and Forte where alt-texts can be added to uploaded images in the Photo app so that they're automatically inserted whenever you embed an image in a message. Using that alt-text field for profile images should be trivial then.

If you want or need the images in my profile described: I don't have a description for my profile photo. For my cover photo, I actually have two. They're both in this post: hub.netzgemeinde.eu/display/c7…. One is in the alt-text which you probably can't open when you're on a phone; the alt-text is 1,500 characters long, a bit over 1,400 of which are visual description. The other one is in the post text itself. But you'd better have a lot of time at your hand because it's over 60,000 characters long, and it'll probably take you a few hours to read it.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Fediverse #Mastodon #Hubzilla #MastodonCulture #FediverseCulture #MastodonCentricity #MastodonNormativity