AI is quite the hot button subject right now. From the positives: summarizing tedious work documents, creating power points, create workout routines or meal plans, story boarding, etc. To the deeper questions: how badly is it impacting the environment, how can I lay off all of the humans and automate every process, how do we get ahead of AI’s impact on art, is it okay to use it to create code?
I want to talk about an interesting side of the AI discussion that I don’t necessarily hear mentioned as much (well, depending on the circle): AI’s impact on the blind community. (Spoiler: it’s still quite contentious.) As a blind person and creator, I find myself deeply conflicted.
How do blind people use AI?
For many, AI has been revolutionary. On apps like Be My Eyes—where blind users can contact a volunteer from anywhere around the world to help them with visual problems, from finding a particular building to identifying a product label—there is an AI feature that will describe any picture you take. You can ask it additional questions afterward to get as much detail as you want, including taking additional pictures from different angles to get more information or if the first photo was not clear enough. There are also the Ray-Ban Meta glasses, which provide AI descriptions of whatever you (and consequently the glasses) are looking at. I have heard of people using the live feature on generative AI models such as Chat GPT and Gemini in the same way as the Ray-Bans, to give live feedback of the environment. The live feature and glasses might be used to describe what’s around you, read mail or books, pinpoint where the entrance to a building is or, very contentiously, help identify when it is safe to cross a road.
Outside of daily living activities, Ai can also be used for describing styles of clothing, celebrities, cover art for things like books, albums or board or card games. There is even the websites ViddyScribe which allows users to produce on demand audio descriptions of videos. AI can describe locations we’ve never visited and would have no way to experience in the same way that a sighted could from a photo or explain visual concepts like a stained glass window. (On that last one, I have seen the pretty colors but never knew until someone told me that they can also have images on them.) AI can also explain how to accomplish skills like styling your hair in particular ways or perfecting a cooking technique that many videos might only express through a combination of verbal explanation and “I’ll just show you” when at a loss for words.
I had an old friend reach out to me a year or so ago, raving about how much of a game changer AI is for blind writers. As I noted above, we can’t get the sense of a new city by going on Google street view and virtually strolling around but we could ask an AI to do it. We could have it describe the architecture, and then have it further explain what those words actually mean. What exactly makes a Gothic cathedral? Could someone really recognize it on sight as compared to brutalist or modernist ones? Where someone might create a Pinterest vision board of what their character wears or how their room or house is laid out, a blind or low vision person could discuss with an AI all of their ideas until they’ve reached a cohesive idea.
Some people will say that you don’t need all of that detail, just the suggestion of a thing is enough. But for some blind authors, we might not have enough of the sense in our own minds to suggest it to someone else. Also, for some people, a few vague terms would be enough as it wouldn’t be the point of the story. For others, they want to understand as much as they can to visualize or conceptualize it into an entity that they can then comfortably write about in their own words.
Where do I stand on the AI debate?
In a maelstrom of confusion.
I have used AI for many of the uses described above. As someone who hates the modern trend toward minimal descriptions in books, I have asked it all of my inane questions about outfits and body types, how to compliment certain color shades that I may not have seen as often so can’t immediately bring to mind. I have also asked tons of non-visual questions. I’ve had it help me figure out a study schedule for brushing up on and hopefully gaining fluency in French. I have asked for the plain English version of medical notes, shortcut the process of determining whether a particular budget would work for Matt and my honey moon plans and countless other things. But after I learned of the various environmental reports, I stopped using the generative models for a while. Then realized that I was actually still utilizing it without think when I used apps like Be My Eyes. So I found myself reopening the other apps with guilt-tinged queries.
Before AI I would ask the people in my life for descriptions and the answers I received would depend on how good or comfortable they were at describing things. How good was their recall? I sometimes found myself holding back or hesitating if it felt like I was asking too many questions. But I couldn’t where out an AI. Except the models also hallucinate. (I actually tried having Gemini Live describe an ice dance performance because it did not have an audio description track and it just began making stuff up, about the movements and clothing. Initially, it seemed accurate enough. Maybe it couldn’t capture all of the quick movements and that’s why it wasn’t actually describing everything with complete accuracy. But then it told me about movements that absolutely had not and were not happening and Matt realized I’d forgotten to turn the camera on.) I also know that nothing beats my senses and intuition, a food thermometer, a Google search where I determine and read through my own sources or just an actually engaging interaction with fellow humans.
I have seen counter discussions about how (some) companies are trying to mitigate the impact. Gemini supposedly routes queries to more available data centers to reduce its impact and improve efficiency. There are other green AI models, both from developers trying to build sustainably from the ground up or platforms that donate to environmentalist groups with each query. This is all helpful and heartening to see.
But I worry about people who ignore the AI’s ability to hallucinate and take everything it says as gospel. Or those who depend on it so heavily they use it for potentially life-threatening tasks like crossing the street. What happens if you start putting more trust in the AI analysis than your own? This is one of the aspects that worry many in the VI community. Also that people will stop trying to make things accessible and just slap an AI bot on a digital product and call it a day. Human reasoning is still integral to this puzzle.
It’s tempting and so seemingly simple to just have all of the information neatly synthesized at your fingertips. I try to make sure that, after I’ve received that plain English explanation, I look up what I understand to make sure it actually aligns, for example, with what I’m seeing or experiencing myself. I also feel better about metrics such as a medium gemini query is comparable to that of a Google search. I do the latter all the time. But I also remain conscious that these are all self reports.
So where does that leave me?
Still conflicted. Feeling a bit less bad but also trying to reduce my initial excited usage. In essence, still figuring out where the line is between the kid who asked her sister if we could completely go green and being told “… that costs money.” And a person, blind or otherwise, who has experienced how useful a tool AI can be.
Be the first to reply