Firefox is adding on-device AI alt text for images.
Generally assistive technology is of course a good thing, however:
How accurate is it?
How accurate can it even be if it doesn’t know the intentions of the poster? (The exact same image can be illustrative to diametrically opposite opinions.)
Where does the training data come from?
Will it discourage authors from adding alt text?
Is there telemetry that could be used to invade privacy? What if that data gets sold or stolen?
@SamStrong if it’s e.g. on social media the intention is very important, the poster likely wants to convey something. Can an AI say an image is “cute” or that it “shows the cruelty of war” or even that it’s “a screenshot of an eBay listing”? Probably not.
@thomasfuchs "In other words - and this is the rock-solid principle on which the whole of the Corporation's Galaxywide success is founded - their fundamental design flaws are completely hidden by their superficial design flaws." - Douglas Adams