Speaker from Totes-didn’t-used-to-do-evil company.
Focus on helping those of us who can’t see very well, or at all.
Lots of discussion of masking things in unicode to try to lure people into visiting bad sites.
I think there might be potential for doing things like confidence intervals, and requirements surrounding the levels required for browsing/redirection. So, the speech to text hit on a potentially-malicious return. The speech-to-text might think it’s 100% confident that that’s what the user wanted.
But you look at the actual amount of traffic to that site, you can say, no, that weird unicode look-alike isn’t what the user was trying to get to.
Were you trying to get to Google, gee-ooh-ooh-gee-ell-eee-dot-com? If yes, hit, “go.” If not, hit “stop.”
For my own stuff, I’m teetering on the edge of legal-blindness. I think last check, I was something like 20/70 in my right (and previously non-dominant) eye, and uncorrectable to 20/400 in my left. I still can type, but some of the predictive things of things like SMS on my iPhone are very beneficial to me.
If I’m not sure, I use a search engine (rarely the totes-didn’t-used-to-do-evil one the speaker worked for….I would say that I’m mostly DDG, with some Bing, and a smattering of Brave), and try to get to the best result.
I do see well enough to do that. But even if I didn’t, I still think there’d be a good way to answer a series of binary questions to get me to where I actually wanted to go.