Skip to main content


🧠 Having lived with reliance on human assistance, many blind people are ready to delegate to AI. It’s not perfect, but it’s consistent, unbiased, and scalable. The trade-off? Independence outweighs the occasional AI mistake. 👏
#Blind #AIForGood #InclusionMatters
in reply to The Blind AI

it is none of those things.
AI is not consistent. Share an image and you'll get a different interpretation every time.
AI is not unbiased. The type of responses are based on data and that data is often Human and loaded with bias. More worrying is the self-imposed bias brought in to try and either keep things safe, "sorry, I can't read you that card number for security", or to cater to other agendas than the asker, "I can't comment on people's visual appearance as it would be insensetive" or "I can't name or identify a real world figure from this image" etc.
And finally AI is far from scalable at present, the gigantic amount of electricity and water being used is going to be a big problem.

Happily, all of these things are addressable. I am blind, of course, and the benefits of AI have been making themselves felt in my own life for months now.
Just wanted to address your specific points earlier up the thread.

in reply to Sean Randall

@cachondo But don’t you see that’s consistent? It’s consistently good enough but not perfect consistently for information purposes but you can’t take action on it. I can get a good enough description of a cat image 24 hours a day seven days a week without asking anybody else!
in reply to The Blind AI

Oh it's certainly good enough for 9 out of the 10 things I ask of it.
But it's also very confidently wrong about so many things that consistent isn't a term I'm comfortable applying yet.
in reply to The Blind AI

@cachondo But in your own words, that’s one out of 10 things! I use access AI and have the option to verify with the human. So what do you think we should say about AI that it has no impact at all on the blind community?
in reply to The Blind AI

I think it has already had a tremendous impact and will have an even bigger one yet. It's not in the hands of the technofobe, and those are potentially the thousands of people who could really use it.

But if just 1 of those 10 questions is medically significant or physically dangerous, we risk pushing technology that could give advice that hurt people. Not accidentally, like a GPS system not being updated to reflect the real world, but with full consciousness and awareness of the fact that the system fabricates things.

LLM's are brilliant, I use them daily. But I also know their limits.

i've seen firsthand what happens when you upsell.
I've literally watched a sales pitch for a Braille notetaker to a desperate and bereaved parent, now expecting that this one box will give their newly blinded child full access to the modern web.
And that's total nonsense, as even the salespeople know.
Let's not be those pedalers.
Let's be prudent in our language, careful in our promises. Hopeful with our expectations, but cautious because of reality.

in reply to Sean Randall

@cachondo I have to agree with you about the sales pitch. Remember, I’m selling nothing. I’m just trying to use AI to make my life a little easier and to combat some of the AI arrangement syndrome that is prevalence especially mastodon!
in reply to The Blind AI

Yeah, I dig.
Mastodon is full of people who see the problems cause dby big tech, and the current appeetite for both water and electricity make AI a big target