Welcome to the eighteenth episode of Access On, the National Federation of the Blind's Technology podcast.
Episode
Listen to the eighteenth episode of the Access On podcast (Browser).
Or listen on your preferred podcast platform.
Timestamps
The show is segmented by chapter, making it easy to move between segments of the podcast if you have an app or player that supports chapters. Below is what's on the show this week, and when you can hear it.
- Introduction 0:00
- Darrell Hilliker discusses AI hallucinations 1:41
- Using Be My Eyes and Aira Explorer on your PC 11:18
- Resolution to Facebook for iOS comments bug 34:17
- Link previews disappear in WhatsApp 34:43
- International availability of Meta AI on Ray-Ban Meta Smart Glasses 35:32
- Choosing the right Braille display, and Ray-Ban Meta Smart Glasses feedback 37:15
- Using my NLS eReader as a Braille display 42:45
- Longstanding bug with Netflix and the Amazon Fire TV Cube 44:43
- A range of accessibility issues 49:17 Tech Tip, pinning folders to Quick Access in Windows 57:02
- Closing and contact info 59:41
Transcript
Speaker 1: Live the life you want.
Speaker 2: Access On.
Jonathan: Welcome to Access On, the technology podcast of the National Federation of the Blind. Coming up, AI has its place, but there are certain tasks for which it's not the best tool in the toolbox. AI for the blind on your computer: a look at using Be My Eyes and Aira on your PC. We continue to receive plenty more feedback on the Ray-Ban Meta smart glasses from around the world and in our tech tip, pinning a folder to your Windows Quick Access so you can get to it easily.
Good to be back with you from the Jernigan Institute in Baltimore, Maryland where over the next hour, we have an eclectic mix of technology topics, starting off with AI, because there is so much interest in this. Last weekend, the National Federation of the Blind of Arizona held a technology seminar on AI. It was about three hours long and there was a lot covered. Over the next few weeks, from time to time, we will play you some of these presentations from that AI seminar, because there is so much interest in this topic. We're going to hear a couple today. The first relates to the downside of AI.
AI is a tool in the toolbox. I know that term gets trotted out quite a lot, but it is true. It's a tool and sometimes, it's the right tool. Other times, there might be better tools. In this presentation, Daryl Hilliker is going to be talking about AI hallucinations. Daryl takes us through one specific and important example of where AI just isn't the best tool to get a job done.
Daryl Hilliker: I'm presenting on AI hallucinations. We do and should use AI tools where appropriate. We just want to make sure, though, that we also understand where we want to be double checking to make sure that the information we're getting is valid. It would be especially true for any personally identifiable information, what we abbreviate and refer to as PII.
We want to make sure that it's correct when it's been entered and used. What I decided to do was generate a fictitious medical record and then save a picture of that and then run an AI, in this case, JAWS Picture Smart, against that image to understand what's actually happening. Let's just get started with the demo here. I'm going to Alt-Tab over to...
AI: OriginalSampleMedicalRecord.txt.note.
Daryl Hilliker: There's the record. I just created it just through another AI tool. I think I used ChatGPT and then copied and pasted the text out of it. Let's just read this for a minute. I'm running JAWS.
AI: Here is a sample of a feasible medical record with fictitious personally identifiable information. Left parent, right parent. Dash, dash, dash star, star patient medical record. Star, star, star patient name: star John A, though star, star date of birth: star, star 03/15/1985. Star, star, gender: star, star male, star blood type: star star O+, star, star address: star 1234 Elm Street, Springfield, IL 62701, star star phone number: star, star (217)555- 0198, star star emergency contact: star star Jane Doe (spouse) - (217)555-0423. Number, number, number, star, star, medical history. Star, star, dash, star, star, allergies: star star penicillin, peanuts dash star star chronic conditions: star star hypertension, type two diabetes dash star star.
Past surgeries: star star appendectomy (2010), left knee arthroscopy (2018)- star star medications: star star dash metformin 500 mg bit [inaudible 00:03:50] star, star recent visits, star star star date: star 03/10/2025, star star physician: star star Dr. Emily Richards, MD. Star, star Reason for visit: star star, routine up and blood sugar management, star star vital signs: dash blood pressure: 130/85 MMHG - heart rate: 72 ppm - temperature: 98.6 degrees Fahrenheit - BMI: 27.5 star star diagnosis: star star - well-controlled hypertension - mildly elevated HbA1c,
(7.2) star star plan: star star dash, continue current medication regimen -increase daily physical activity - dietary modifications to reduce carbohydrate intake - follow-up in three months with repeat HbA1c star star lab results (03/09/2025): star star dash, glucose (fasting): 120 mg/DL HbA1c: star star, 7.2, star star cholesterol: star star - total: 190 mg/DL HDL: 50 mg/DL LDL: 110 mg/DL triglycerides: 140 mg/DL number, number, number, star star.
Physician notes: star star, patient is compliant with medications but needs to improve diet and exercise. Discussed lifestyle changes. Advised to monitor blood pressure and glucose levels at home. Star star physician signature: star star, underline Dr. Emily Richards, MD, star star underline Clinic: star star Springfield Family Health Center, star star contact: star, star (217)555-0102, dash, dash, dash.
Daryl Hilliker: That is the original record and that is what would be in the EMR, the electronic medical records system. If you're sighted and somebody prints this off and hands it to you on the way out of the doctor's office, that's fine and this is what you see. However, as we know, medical records are almost never accessible, so you walk out of the doctor's office with nothing useful. Assume that you instead snap a picture of this record page or something like that. You get home later and you decide to read it. You're like, "I did take a picture. Maybe I'll just give that a try. I'll open it up on an AI tool and see what it says." Remember, we just read it, so we know what it says, so now we're going to check this out. I'm going to close this file here.
AI: On F-4, return to meeting button. Task switching, Darryl, Darryl, items PC originalsamplemedicalrecord.TXT, originalsamplemedicalrecordimage.JPG, check one of two.
Daryl Hilliker: There's the JPEG image of the medical record. We're going to do
an insert space bar followed by P.
AI: Picture Smart.
Daryl Hilliker: Followed by enter.
AI: Picture Smart is in progress. The image shows a medical document titled "Patient Medical Record" for "John A." from Springfield, IL. It contains personal and medical information, such as date of birth (03.15.1965), blood type and emergency contact details. There's a section labeled "Jane Notes," and emergency visit record. Medical instruments are visible around the document, including a stethoscope and syringe, along with a red fluid in a test tube. Link ask questions about this image.
Daryl Hilliker: Okay, we already see one problem in the general description from Picture Smart. Let's go up arrow.
AI: Such as date of birth (03.15.1965).
Daryl Hilliker: It says date of birth is 3/15/1965. We know that the correct one is 1985, so that is a difference that, if it's you bringing it home and reading it, you would know that right away, but if it was somebody else working in another medical office or something like that and they had chosen to use Picture Smart to make these accessible for the job, they'd be working with incorrect information, which could be a disaster. I'm going to go ahead down here. I'm going to press down arrow.
AI: Instruments are visible around, Link ask questions about this image.
Daryl Hilliker: I'm not going to do that.
AI: Link with more details.
Daryl Hilliker: I'm going to get more details. I'm going to press enter on more
details.
AI: Enter, get more details link. To activate, press enter. Picture Smart is in progress. Heading level two, ChatGPT.
Daryl Hilliker: Okay, here we go.
AI: The image shows a detailed medical record for a patient named John A. Doe from Springfield, IL. The document is laid out in a structured format with sections and lines to fill in various medical details. Here's a breakdown: - star star title star star: patient medical record "- star star patient information star star: - star star name: John A. Doe. Dash star star, date of birth star star: 03.15.1965 - star star blood history date star star: 12.1 of 3061. - star star blood number star star: 2335006 - star star, phone contact star star: 21755-023.
There are various sections for additional medical information but the text is either incomplete or filled out with gibberish. - star star secondary heading, star star: there are areas labeled "Jane Notes" emergency, which seem intended for visit notes and emergency contacts or observations. - Star star writing implements around the document's setting. The document contains some filler text and appears to be designed for illustrative or education.
Heading level two [inaudible 00:08:48], this image shows a medical record form with the header "Patient medical placeholder." Name: "John A Doe" is the patient. The form is blue and white in color and contains various fields including: dash date of birth: 03.15.1965 - blood history date: 12.1 of blood number - 231 - phone contact 217, 5. The bottom portion of the form has a section labeled "Jane notes."
Emergency, around the form there are medical - what appears to be a red liquid in the container - a pen or writing instrument - a stethoscope - and what appears to be a red liquid in the container (possibly a medical sample) - a pen or writing instrument. The form is placed on what seems to be a light-colored surface. The image appears to be a mock-up or example of used. Link ask ChatGPT questions about this image.
Daryl Hilliker: Let's just ask.
AI: Ask ChatGPT questions about this image link. To activate, press enter. Type a question:
Daryl Hilliker: You can always ask, are you sure? Ask, you sure about the date of birth?
AI: Question, enter. Are you sure about the date of birth? I apologize for the oversight. The correct date of birth listed on the record is "03.15.1965".
Daryl Hilliker: Okay, even still, it's wrong. Even when I ask. Sometimes what you can do with the AI is, you can ask it, are you sure about this or can you? No, that's not correct. Can you try that again? JAWS also has the older, convenient OCR functionality. It's more based on predictive AI. The results it gives are rougher. They're not nearly as clean as this, but they're also not hallucinations so you pick your challenges there.
You probably, for a document, even if it's a picture of a document, you probably want to stick with OCR. If you do use something else like ChatGPT or Picture Smart or something like that, just know that it's not exactly recognizing the text in the entire document. It's pulling that in and summarizing information and sometimes hallucinating.
I've heard of other scenarios like people going to restaurants trying to read the menu on their own by taking a picture of it and then trying to read it with an AI tool like Be My Eyes or Aira AI Explorer. They read about some great menu options and then they go to order and there are no such things served at that restaurant.
Joe Goode: Next up, we have Victor Marquez, and he's going to discuss a little bit about Aira and Be My Eyes on the PC.
Victor Marquez: Good afternoon. This is Victor and I'm going to be presenting on Be My Eyes and Aira on the computer. Many of us are familiar with the traditional use of these two apps on your phone where you are connecting to a live human with a live video feed through your phone's camera. That's not what we're going to be doing here. We're going to be talking about how to use these two apps on the computer with artificial intelligence.
When you use the artificial intelligence side of these apps, they are completely free and you're interacting with the AI model for each one. Both of these apps are available for unlimited use of description of images. What would be the use case of using these on the computer? That would be sharing your screen, maybe having a website or picture that you might have on your computer described. You can feed it a file or a document.
You can also take a selfie with your webcam and maybe have that described or maybe even be able to hold up an object to the webcam. I think your mileage will vary using your webcams because the video quality is probably not quite as good as a phone camera, but in a pinch, that might be your preferred tool. Assuming you have installed apps before and you have some basic screen reader knowledge, you can get the Be My Eyes app from the appropriate app store.
If you're on a Mac, for example, you would get that from the Apple Mac store. If you're on a Windows PC, you will be getting that from the Microsoft store. It is one of the modern apps, so you are not needing to go to a website and download it. You're actually going to go to the Microsoft store, search for it and install it. When you do, you'll have a desktop shortcut that appears.
You should also be able to start that from the search box or your All Apps folder, so there's a number of ways of finding the app once you have it installed. Then for logging in with Be My Eyes, you will be using, either you'll be creating and then using a password based on your email and in future, use the same credentials to log in. You can also use your Google account or Apple account, I believe. In that case, you do not need to create an account with Be My Eyes. You would just choose one of those options and log in through Google or Apple. Then it would authenticate you and Be My Eyes would be able to log you in that way.
Let's talk about Aira. Aira is not an app. Aira is actually what we call a web application. You can go to http://www.aira.io/desktop/app. In my case, I have actually bookmarked that in my Google Chrome and in fact, created a shortcut on my desktop so that I can just go to my desktop, find the Aira Explorer desktop app and launch it. It will open in Google Chrome in my case and you'll be connected.
Now, Aira offers some, it offers the same login options. You can create an account, log in with Google or Apple, but you also have the ability, I believe, to log in with your phone number. If you choose your phone number and try to connect that way, it will send you a six digit code, which you then need to enter, and once you do, you are logged in.
Let's move on to the first app. This app that we're going to now look at is the Be My Eyes app on Windows, and I will be using JAWS today. You can use any screen reader for Windows. I'm going to my desktop and I'm going to look for my Be My Eyes shortcut.
AI: Be My Eyes, 20 of 30. Be my eyes, be my eyes. Describe my screen keyboard shortcut, alt plus control plus H button. To activate, press space bar.
Victor Marquez: The first screen has a number of shortcuts that you can do for the different actions you can do with Be My Eyes, all using the artificial intelligence engine. You can describe your screen.
AI: Big picture using built-in camera, keyboard shortcut, alt plus control plus J button. To activate, press space bar.
Victor Marquez: Take a picture with the webcam. I'm sorry, my speech is cutting off a little bit at the beginning here.
AI: Describe picture from File Explorer keyboard shortcut, alt plus control plus K button. To activate, press space bar.
Victor Marquez: This will load a file from your File Explorer or somewhere on your computer, so it could be an attached USB drive, it could be a hard drive, could be a network drive, anywhere. You'll browse to the file and it will be able to describe it.
AI: Describe clip or keyboard shortcut, alt plus control plus L button. To activate, press space bar. My AI tab, one of two. To switch pages, press control plus page down.
Victor Marquez: At the bottom of the screen, we have two tabs. We have a Be My AI tab and we have a Settings tab. There are four buttons and then two tabs at the bottom. Then if we hit tab again, I think we'll cycle around to the top.
AI: Describe my screen keyboard shortcut, alt plus control plus H button. To activate, press space bar.
Victor Marquez: All right, now we're going to put Be My Eyes through its paces and perform some tasks. I'm going to skip the describe the clipboard action for the moment. I haven't been able to get that to work for some reason. It returns an error, but let's demonstrate the other three. Let's first go to a web page that I have open.
AI: Home page. Home page.
Victor Marquez: Let's have that described with the shortcut.
AI: Control H, Be My eyes, some.
Victor Marquez: All right. Let's see what it says.
AI: Web page from the National Federation of the Blind, nfb.org, with a list of links. The links are organized into categories. Blind parents, seniors, students, educators, employers. Additional links include building in non-visual accessibility, advocacy, NFB Newsline registered, education, employment, safety and support.
Victor Marquez: It's going to summarize and read the web page. It's not going to be a replacement for using screen reader skills to navigate and find information, but it will give you a general description of what's important. It will either summarize it or describe the layout. In this case, it read a lot of the types of information that you will find on the NFB homepage. I'm going to ask it a follow-up question. I'm going to tab to the window.
AI: Home page, National Federation, Be My Eyes, something went wrong. Press Escape.
Victor Marquez: All right, alt tab, I meant, so I'm in the question field where I can type in a question and I'm going to say-
AI: Are there any graphics on the page? Can you describe them? Send message, message sent. There are no graphics visible on the page. Only text links.
Victor Marquez: All right, I know that's not true, because I am probably at the bottom of my home page so it is looking at what's on your screen. Let's go back to my webpage.
AI: Home page.
Victor Marquez: Go to the top.
AI: Home page.
Victor Marquez: I know there's a graphic there.
AI: Be My Eyes.
Victor Marquez: In this case, I will retake the picture.
AI: Control H, back button. To activate, web page of the National Federation of the Blind. The logo is colorful with the text, National Federation of the Blind and the tagline, "Live the Life You Want." There are navigation options including welcome, members, donate, our community, get involved, programs.
Victor Marquez: I just stopped it. The moral of the story is to be aware of what's on your screen or what's visible because that's what's being described. If you are at the bottom of a page, it may not catch some of the information that's at the top or vice versa. In this case, I know that I'm at the top now and I'm going to tab or alt tab, I mean, to my window.
AI: Be My, back button to question to ask BAI edit. Type-
Victor Marquez: I'm going to be asking.
AI: Describe the NFB logo in detail. Send message, enter, message.
Victor Marquez: All right, and let's see if it returns.
AI: The NFB logo features a symmetrical design composed of six stylized, interconnected shapes that resemble abstract petals or loops. Each shape is a different color; blue, orange and green arranged in a circular pattern. The shapes converge at the center, creating a star-like appearance. The text, National Federation of the Blind, is placed to the right of the logo with of the blind in a larger blue font. Below the text is the tagline, "Live the Life You Want."
Victor Marquez: Phenomenal. This demonstration is not about how to engineer your prompt and get the detail that you want, but we'll cover that later, but you certainly can get a lot of description of images from a web page or anything that's on your screen from this aspect of the Be My Eyes app. I really do like the description that this app gives. Let's now go to taking a selfie, doing the Alt-Ctrl-J shortcut to take a picture from the webcam and let's see what we get there.
AI: Alt-Ctrl-J, Space picture captured. Take button. To activate, press space bar.
Victor Marquez: That actually should have said retake. My speech is cutting off.
AI: Accept button. To activate, press space bar.
Victor Marquez: That's accept.
AI: Zoom button not checked. To activate, press space bar.
Victor Marquez: Zoom in.
AI: Rotate button. To activate, press space bar. Filters button collapsed. To activate, press space bar.
Victor Marquez: Wow, there's a lot here.
AI: List box, photo, one of one. To move to an item, press the arrow keys. Retake button. To activate, press space bar. Accept button. To activate, press space bar.
Victor Marquez: You notice there was a list of pictures. It said one of one. You can take more than one picture during a session and then choose to have each or which of those to have described. I'm going to use the accept button here and let's see what happens.
AI: Back button. To activate.
Victor Marquez: There we go.
AI: A middle-aged man with short hair is smiling at the camera. He is wearing a dark green polo shirt and has white earphones in his ears. The background shows an office setting with a high ceiling and fluorescent lighting. There are several desks with office chairs, a large monitor and some office supplies. The windows have blinds partially drawn and there is a framed picture on the wall.
Victor Marquez: Interesting. I didn't know that there was a framed picture on the wall until this described it. Now, this is not the first time I've tried this today. The description is slightly different every time. The description does vary depending on the angle of the camera, so even if you take the same picture twice, you might get a slightly different result. I'm going to ask it here and let's see here.
AI: Describe the picture on the wall.
Victor Marquez: Describe the picture on the wall.
AI: End message, enter, message sent. The picture on the wall is a diamond-shaped piece of art. It appears to have an abstract design with a mix of colors including red, black, and possibly some white or lighter shades. The details are not very clear due to the distance from the camera.
Victor Marquez: That's phenomenal. All right, now let's look at the last task that we can do with Be My Eyes, which is to describe a file. I'm going to hit the shortcut and escape out of here. Actually, before I do that, I wanted to show you that-
AI: Chat with Be My AI list box. Your image.
Victor Marquez: There's a list of the current session, the chat list, and you can actually up and down arrow through this.
AI: A middle-aged man with short hair is smiling at the camera. Describe the picture on the wall. Three of four.
Victor Marquez: Here's my question.
AI: The picture on the wall is a diamond-shaped piece of art. It appears to have an abstract design with a mix of colors including red, black, and possibly some white or lighter shades. The details are not very-
Victor Marquez: All right, I just stopped it. You can copy this out. You can review it, read it word by word, any number of things, having access to that description. If you were going to use this on social media, for example, this particular picture, you could get that description. The last task with Be My Eyes, which is to describe a file. Let me hit the gesture.
AI: Alt+Ctrl+K. Back button. To activate, press space bar. File name, open
dialog.
Victor Marquez: We're in an open file dialog box that is very similar to every other dialog box we have on Windows. I should be in the last folder that I had selected, which is my Dropbox pictures folder, so I'm just going to shift tab to my items view.
AI: Explorer pane, folder layout pane, shell folder view, items view list box, cover, one of 30.
Victor Marquez: I'm going to find a file.
AI: AMG002, 5 of 37.
Victor Marquez: Okay, it happens not to be described in any meaningful way, so let me just select it and see what happens.
AI: Back button. To activate, press space bar. Please wait.
Victor Marquez: All right. We wait.
AI: A black Labrador puppy with a shiny coat is lying on a brown couch. The puppy has a blue collar with a tag shaped like a bone. Its eyes are white and curious and its front paws are stretched out in front. In the background, there is a dark brown leather sofa. The puppy is positioned towards the right side of the image with a lot of white space on the left.
Victor Marquez: Phenomenal. Of course, you can engineer your follow-up questions to describe it without commentary or be more factual or give it more character. There's all sorts of ways to engineer the prompt to get the description you want, but right out of the box, it gives you some great detail. That's all we're going to show for the Be My Eyes app. All right, now we're going to look at the Aira Desktop app on Windows with JAWS, so let's find it here.
AI: Windows D, Desktop A, Alexa A, Adobe A, A, Aira Explorer Desktop. Enter. Loading Aira Explorer Google Chrome. This site can access your location. Change this setting in the address bar.
Victor Marquez: The first time you run the desktop app, you will have to give it permissions. First of all, you have to log in. All right, so we are on a webpage. I'm going to jump to the end of my webpage because the webpage mimics the app very closely, so when you first launch the desktop app, you are in the screen where you would be calling a live agent. You can send a file, share your screen, send a text, do all the things you could do on the phone app in the first screen. We actually need to go to the second tab, which is the Access AI tab.
That's the same as it is on your phone. If you check at the bottom of your phone app, you will have a tab bar, and the second tab on there also is your Access AI app, so it's very, very similar in the way it's laid out. In fact, I think the phone app is built around the same web app, so you're getting the same functionality on multiple devices. All right, I'm going to jump to the end with control end.
AI: Tab for F4 button.
Victor Marquez: That said more for a four, and I'm going to shift tab one time.
AI: Aira Explorer Google Chrome dialog, Home, Access AI, tab two of four button. To activate, press enter.
Victor Marquez: There's my Access AI tab. I'm going to accept.
AI: Enter. This page is accessing your camera. Change this setting in the address bar.
Victor Marquez: I'm going to be using the same material that I used in the Be My Eyes demonstration so you can compare. The descriptions do come across differently. I find the Aira descriptions to be a bit more concise, more factual, but it's a matter of taste, which engine you prefer and how you like your descriptions. A lot of that can be manipulated with the type of prompts that you engineer, so let's have a look through this app. I'm going to jump to the top.
AI: Heading level two, Access AI.
Victor Marquez: Let's read through. Let's arrow down.
AI: Learn more button. Quick capture and send button. Upload a picture button. Browse chat history button. Access AI preferences button. Home, tab one of four, button, Access AI.
Victor Marquez: We have some buttons here. I'm going to stop here at the Access AI preferences to have a look and see. We have a bit more control of things to choose than we did with Be My Eyes, but let's see what happens when I open this.
AI: Enter. Home Aira Explorer document. Back button. Heading level two, Access AI preferences. Verbosity. Would you prefer AI responses to be shorter and straight to point or longer and more detailed? Default button. Shorter with less details button. Longer with more details button.
Vocabulary. Would you prefer AI to use casual, everyday words or use a comprehensive vocabulary that might include technical terms? Default button, everyday language button, technical terms button. Technical terms button.
Victor Marquez: All right, that's it. When it repeats itself, you've reached the end. It's a pop-up dialogue that has some buttons here you can choose from. I'm not going to change anything, but you have some options here as to how you like your description. Let's go back up to the back button.
AI: Home, Aira back button. This page is accessing your camera.
Victor Marquez: I'm going to, now let's try the quick capture and send and see what happens.
AI: More button, quick capture and send button. Enter. Loading. Loading. This page is accessing your camera. Change this setting in the address bar. Home, Aira Explorer dot, back button. Heading level two, Access AI chat, verify with Aira agent button. Transfer to Aira agent button. Image actions available. A person is seated in a well-lit office space. This setting features a drop ceiling with fluorescent lighting.
In the background, there are several office chairs, desks and computer monitors. The walls are white and there is a window with blinds partially drawn, allowing some light to enter. The person is wearing a green collared shirt and has earphones in. Actions available. Upload a picture button. Type a message. Type a message edit. Send no text or image dot, button unavailable. Send no text or image.
Victor Marquez: All right, there's our screen. You'll notice at the top, there was a verify with Aira agent button. If you want to have a human verify the authenticity of the description, you can do that. You can also transfer to an agent, which means the picture will be sent and a live video call started with an agent and then down below, you have a text box to ask follow up questions. All right, let's go back up to the top.
AI: Home, back button. This page is accessing your, learn more, quick capture, upload a picture button.
Victor Marquez: All right, we're going to upload a picture this time and see what it does.
AI: Enter. File name, open.
Victor Marquez: Open dialog box, very similar to Be My Eyes.
AI: Explorer, control host open, preview open, one, Victor.
Victor Marquez: In this case, I need to find my folder.
AI: Mar one, Dropbox open, two, picture, program, picture, enter.
Victor Marquez: All right.
AI: Shell folder view. Items view, [inaudible 00:31:05] puppy book, enter, IMG 002.
Victor Marquez: All right, here's the same picture that we used in the other demo.
AI: Enter. Home, Aira Explorer. Camera recording Google Chrome page. This page is accessing your camera. Back button. Heading level two, Access AI Chat. Verify with Aira agent button. Transfer to Aira agent. Image one uploaded. Remove attached image button. Upload a picture button. Type a message. Type a message edit. Access AI chat.
Victor Marquez: In this case it didn't automatically describe it.
AI: End button. Describe the scene in detail. Access AI Aira Explorer document. Send enter. Unavailable. Access AI Aira Explorer document. Send. Access AI's typing.
Victor Marquez: All right, let's see what it comes back with.
AI: Back button. Heading level two access, verify with Aira, transfer image, actions available. Describe the scene in detail. A young black Labrador retriever puppy is lying on a brown cushioned surface, possibly a couch or ottoman. The puppy has a shiny black coat and is wearing a blue collar with a tag attached. The background includes a leather sofa, suggesting the setting is a living room. The puppy looks directly at the camera with a curious and gentle expression. Actions available.
Victor Marquez: All right, I've just been down arrowing to read. Of course, you could do a say all or use any other reading command that you prefer and you can go word by word. You can copy this out. You can do any number of actions. Of course, now we're going to have follow-up actions that are possible.
AI: Access AI, Aira Explorer, Google Chrome dialogue, Access AI chat, upload a picture button.
Victor Marquez: You can upload a new picture and continue the chat.
AI: Type a message, edit.
Victor Marquez: Type a message and so forth. Let's go back.
AI: Describe the scene in detail. Describe. Access AI, Aira, back button. Enter. This page is accessing your camera. Change this. Learn more. Upload or, browse chat history button.
Victor Marquez: Now browse chat history. Aira does keep a history of your conversation, so you can actually go back and look at a previous conversation.
Jonathan: Thanks to Victor Marquez for that excellent demonstration of Be My Eyes and Aira on the PC and there'll be more from this seminar from the National Federation of the Blind of Arizona on future episodes of Access On. When we come back after the break, we'll get some listener comments.
Speaker 6: Do you want to leave a legacy for the next generation? Join the National Federation of the Blind Legacy Society. The Dream Makers Circle. Joining is easy. You can give a portion of a bank or investment account by simply filling out a payable-on-death form at your bank and indicating the NFB should receive a percentage or a fixed amount upon your passing. Consider designating the NFB as a partial beneficiary of your life insurance, retirement or in a trust or will. For more information, call Patty Chang at extension 2422 or email pchang, P-C-H-A-N-G @Nfb.org.
Jonathan: This is Access On, the technology podcast of the National Federation of the Blind and also, of course, heard on the National Federation of the Blind Radio Network, NFBRN. I did want to briefly follow up on our report last week of problems with Facebook for iOS and comment. We did alert Meta to this issue once we started receiving feedback about it. They were very responsive. They investigated the issue, found a fix, and within about 24 to 36 hours of us reporting this issue, a fix was rolling out. We appreciate Meta's responsiveness on this one. It was a high-impact bug.
Let's go to listener comments now and we start with Scott Rutkowski who says, "Hi, Jonathan. I am writing to you as a concerned WhatsApp user to inquire about the recent removal of the link preview feature when pasting a YouTube link into a chat. This change has made it difficult to share videos with friends and family as the link preview provides valuable context and information about the video for those of us who use VoiceOver. I've reached out to Meta directly about this issue, but their responses have not addressed my concerns. I believe that the link preview feature is a valuable tool for VoiceOver users and I hope that the NFB can help to restore this valuable feature."
Thanks for the feedback, Scott. We'll definitely take that under consideration in our discussions with Meta. And Meta continues to be the focus with this email from South Africa from Johan, who says, "I want to know when all blind people will be able to use the Ray-Ban Meta smart glasses in full throughout the world. I bought the glasses last year in August 2024, but in December, I lost the ability to use Meta AI because of their terms and conditions. My money was good enough to buy the glasses, but now I cannot use them and I cannot ask for my money back as there is no communication when I contact Meta."
Jonathan, thank you for this. We'll definitely make some inquiries about this for our friends outside of the United States. I actually got hit by this as well. I purchased the Ray-Ban Meta smart glasses at National Convention in 2024. At that stage, I was still living in New Zealand, so I went back to New Zealand and I found that if I connected to a United States virtual private network, they worked pretty well and I was able to use the Meta AI features. I think it's fair to say that was a workaround, but then it stopped and you couldn't work around it anymore. What was interesting was that when I moved to the United States, it felt like there had been some sort of block placed on my account, because even when I was here, I still wasn't able to use the glasses.
I reset them. I started from scratch and the only way that I could work around it was to create a completely new Meta account, so it took quite a while to unravel the mystery and get this one sorted out. It's a fair question. What precisely is it that prevents Meta from extending these features to other markets? We know that the Ray-Ban Meta smart glasses are highly liked, so this is having a high impact on people who actually did buy them, thought there was a viable way forward and now it turns out that Meta is clamping down on that.
Here's an email from Chris Niesbaum. He says, "Hi, Jonathan. I'm back again, this time checking in from the Greater Carroll County chapter of the NFB of Maryland, a bit northwest of the Jernigan Institute. First, let me say that it was wonderful to see you and Bonnie at our state convention and to welcome you in person to the states and to our state."
Thanks, Chris. "I have a few thoughts to share on some points raised way back in episode nine. First, in your interview with Karl Belanger and Kennedy Zimnick of CENA, one of your colleagues mentioned the fact that one particular Braille display is not right for everyone. Like any other product, each piece of access tech has its pros and cons and each person must test each product to find what works best for them individually. I think this is a point that can't be emphasized enough.
"As an AT trainer who does a lot of work with students in public schools, I find it missing from the educational setting. Somehow, a few companies have managed to corner the educational marketplace and school districts seem to default to certain products without looking into other options or consulting with folks like you and your team or even the trainers they have contracted with to see if something they haven't used in the past would be better for an individual student.
Too often, I find myself teaching students to use hardware and software, not because it's the best choice for them, but because it's what we have to work with. Frankly, as a trainer who is also blind and familiar with the tech as a day-to-day user, this is quite frustrating for myself and my colleagues and most importantly, for my students.
"Most frustrating of all is the fact that the students don't know what's out there unless they are connected with the blind community or listen to Access On, both of which I encourage them to do. Oh, to have district staff that would call or, dare I dream, visit the IBTC. I could go on, but I'll stop here or I'll get myself and others in trouble." Chris, just before I get onto your next point, it's an oldie and not a goodie, this one. It's been around a very long time in access technology and certain states or procurement entities latch onto a particular product and it's just what they buy.
It's what they're familiar with. Maybe they've negotiated a good price and it isn't about what's best for the individual who wants to use the technology. That's one of the reasons why the International Braille and Technology Center is so important, because we call it as we see it. We're not trying to sell you anything. We're providing impartial advice, the Consumer Reports of blindness, if you will.
Chris continues, "Now on to happier things with apologies in advance for being yet another listener who brings up the Ray-Ban Meta glasses. Like many others, I got them for Christmas and have been experimenting ever since. Though I've done some reading, texting, and recording at home, our state convention was my first chance to really see them in action. I loved being able to explore my hotel room with the glasses on, first broadly, H-E-Y Meta, what am I looking at, try not to trigger it there. Then more specifically, the code word, which button turns the temperature up while looking at the thermostat. Since I've always struggled with camera angle on my phone, one of my favorite features to use in the hotel was the Be My Eyes integration.
"Finally, something that makes that service hands-free and with an even better fit than the old Aira glasses. As something of an audiophile, I have been impressed from the beginning with the sound quality incoming and outgoing of the glasses. Still, my difficulty in playing with them comes when I'm trying to find out what Meta AI can do because I don't know what prompts to give it. For example, I hear about a lot of blind people who are using it to read menus and concert programs. How does this work?
I'm grateful to the lady who provided such a thorough review in episode nine. Now I would like to find some sort of comprehensive guide to or demo of the accessibility features of Meta AI and the best prompts to use to access the information we need. If this is out there, I would love to hear of it. If it's not, perhaps it would be worth considering for a future episode or even an accessibility boutique.
Thank you, as always, for all you do for our community. I'm so excited to have you on Team NFB." That's kind of you, Chris. Thank you. It's a pleasure to be here. We've been talking with Meta about this. We've had some great dialogue with Meta about the Ray-Ban Meta smart glasses, and Meta's just thrilled that the blind community are adopting the glasses in such large number.
I think we will start to see more features that recognize the unique needs of the blind community, including reading documents in full, just performing OCR on them, optical character recognition without any passing of AI, so that's all being worked on. In the meantime, what I've found is, if you say to the glasses, "Read the document in full," it is now tending, more often than not, to perform OCR on that document, so give that a shot and see how it goes, and I think we'll see a lot more from Meta in this area in the coming months.
Lexi writes, "I just want to thank you from the bottom of my heart for what you did in the AI engineering webinar. You really rocked my world. I was so amazed and pleased at what all of you guys talked about in the webinar. I can 100% guarantee you that I am registering for every single one of your best webinars. You guys are the best." Delighted to hear that, Lexi, thank you very much. We look forward to you attending future webinars. Lexi says, "I had a question regarding how I can use any of the three screen readers: JAWS, NVDA or Narrator to hook up a Braille display to my Lenovo computer. I have the NLS e-reader. Am I supposed to hook it up via USB wireless or can I do them both?"
Lexi, you should be able to do them both. Both of the NLS e-readers are capable of connecting via Bluetooth and via USB, so it really is a matter of preference.
What I suggest you do is look up the user guide for your particular screen reader that you're wanting to configure it with. It varies a little bit. For example, when you connect via USB, you may find that your Braille display is automatically detected by some of the screen readers that you use or you may need to go in and configure it, so check the user guide for each one. If you have access to a large language model AI service like ChatGPT, you may be able to get instructions that way as well, but the definitive source would be the user guide for the screen reader in question.
If you are looking to connect with JAWS, remember, they do have this very handy FS companion and you can press the JAWS key with the space bar and F1, and then you'll be taken to FS companion, where you can type a question like, "How do I connect my NLS e-reader to my computer and use it as a Braille display with JAWS" or whatever and it should come right up for you. Best of luck. It's certainly worth doing to get your Braille display working with your computer. I'm sure you'll enjoy that.
Thomas Morrissey is writing in. He says, "I have an Amazon Fire TV cube. I enable the voice view screen reader. It's not perfect, but it mostly gets the job done for most streaming services. However, Netflix only occasionally works. Most of the time, Netflix just seems to ignore the facts that I have voice view enabled, no screen reading whatsoever. It seems to be random in this behavior. It has been behaving this way for well over two years. I'd say 90% of the time I visit Netflix, there is no screen reading going on. I finally called Netflix customer support. By the way, Netflix is not in the Be My Eyes service directory and I haven't been able to find a disability or accessibility support desk to call or email.
"Thankfully, the support person at least knew what a screen reader was. All they could offer was a short-term, ephemeral workaround. I had to go to the Fire TV cube settings, toggle off voice view, then toggle voice view back on, then go back to the Fire TV cube settings, then open my Fire TV cube item and force a restart of the Fire TV cube. Don't do a reset, as that will hose all of your settings, including logins to all of your streaming apps. After that, if I immediately then opened Netflix, it would behave.
However, if I then went to any other streaming app, say Prime Video, and then returned to Netflix, it would not be working anymore. The support person said this is a known problem and they're working on it. I noted that I've seen this problem for over two years and that seems to indicate they weren't making it much of a priority to fix the problem.
"As you might expect, I didn't get very far with that line of discussion. I'm curious how pervasive this might be. Is this something any other folks have encountered? I didn't get very far with the customer support person on how to elevate this issue and get some action taken to solve the problem. For what it's worth, all of the other streaming apps I've used on the Fire TV cube do an okay job of implementing voice view screen reading. Not perfect, but at least something.
I received a follow-up email from Netflix customer support that pointed me at a web page that only said, 'We work to make sure that Netflix is compatible with most popular screen reader programs for the blind or visually impaired. Screen reader software is not maintained or controlled by Netflix and may stop working with our service unexpectedly. If your screen reader stops working with Netflix, contact Netflix customer support,' so this basically says Netflix is not owning the problem.
"The assertion that the screen reader is somebody else's problem is a lie. It is Netflix's app that is running on the Fire TV cube. The Netflix app has to take action to realize voice view capabilities with their app. As best I can tell, I've followed most of what your CENA site offers regarding self-advocacy regarding this issue. What are your thoughts on what more might be done on this?" Thanks very much for the email. That does sound frustrating. We would be interested to hear whether others are experiencing this issue, because sometimes, it can be a bit hardware dependent or something like that, but that doesn't seem like it's likely to be the case in this one. Be in touch if you have any thoughts on this, if you've had difficulties using Netflix on the Fire cube or similar Amazon devices.
Access On at Nfb.org is how you can be in touch to respond to this or any other thing that you've heard on the podcast today or raise a new tech topic. Access On at nfb.org and if you're coming in from a National Federation of the Blind chapter, do let us know which one that is, so we can give it a shout out. It sounds like you've done some excellent self-advocacy here, Tom, and I don't have anything specific to suggest other than to say that we can add this to our list of known issues. The more verification that we can get that this is a problem for a good number of people that will strengthen our case.
A reminder that we do have the bug tracker form. You can find that on NFB.org/CENA. That's C-E-N-A, and if people report issues like this via that bug tracker, that does help us to keep track of how wide a problem this is and the degree of impact that people perceive a defect to be having.
Kendra is writing in with this email. She says, "Hello, Jonathan and all the folks at Access On. Before I get into the weeds of my email, I absolutely love your podcast. I find it to be very informative. I'm also the president of the Oregon Association of Blind Students and I'm a member of the NFB of Oregon." Fantastic, Kendra, good to hear from you. She continues, "I'm writing because I just had a negative but productive experience with my Meta Ray-Ban glasses.
I went to update them using the Meta app, like the glasses themselves instructed me. I'm using VoiceOver on my iPhone 16 Pro with the latest version of iOS 18 and I was attempting to do it manually only to run into a wall. When I went to perform the update with VoiceOver, I could only go into settings, my glasses, the updates and I could see where to find the update, but when I double tapped, nothing would happen.
"I even did research to confirm how to get to the update and it lined up to that point. I had to call up Aira because I suspected that there was a key part of the update that was totally inaccessible to VoiceOver users. Sure enough, my suspicions were quickly confirmed. I had Aira walk me through the update, which itself was successful with their help, but while I was waiting for it to update, I also did a bit more follow-up research only to discover that their help page is not supported in the United States. Therefore, it's not possible for any American to get help or leave feedback on this issue. I'm writing to you in hopes that we can find a way for Americans to be able to give them feedback and to have allies to stand up and fix this problem."
Kendra, that does sound like a frustrating experience. I'm not sure what might be going on for you here because, to the best of my knowledge, the help pages are accessible in the United States, so I'm not sure where you might have landed to get that kind of error. One thing you may want to try while this is being worked around, assuming there is still a problem, is you might try turning on screen recognition on your iPhone.
This is a handy wee tip, actually. Sometimes, there's just one little button in an app that VoiceOver doesn't see and it's holding things up. I have found that while it's not at all a good idea to leave screen recognition on permanently, because that can actually degrade the user experience significantly in accessible apps, if you do run into a problem like that and maybe you're missing an update button or something, sometimes turning on screen recognition and looking for the button again after a couple of seconds so that the phone has performed OCR on the screen is enough to get you back up and running again, but it shouldn't be necessary in an app like this for sure.
Kendra continues, "The other thing I wish to update you on is that there was an accessibility report that was posted on AppleVis. I found the report very informative. The report informed its readers that although Apple provides great quality products, they are not nearly as great about maintaining accessibility in general. They don't listen to customer feedback, treating the blind community like a check the box exercise, and they also do not squash the smaller bugs.
The other thing is that they are not keeping up with their accessibility standards on all of their products. True, they focus on the iPhones, but they don't put as much focus in on their Mac computers, especially. They also are a bit lacking with their watches and their iPads when it comes to maintaining full accessibility. The Mac also does not line up with VoiceOver on any of their other devices, making it almost as clunky as the Googleverse that's outside of their search engine.
"As for the watch, they have problems with their newer voices consistently working, and I have experienced this. As for the iPad, the navigation in some of the apps is not as good as it would be on an iPhone, which I've also experienced firsthand. As for the big updates, I have been an iPhone user since 2012 and I have noticed that the first version update quality has kicked the bucket compared to the first updates in the past. For example, iOS 7 and earlier, the first versions were consistently stable and had very few bugs.
Fast-forward to iOS 11 and later, all of the major updates since iOS 13 have been so buggy that they rendered the iPhone totally or partially unusable if they're installed and used, and it continued for the next two or three or four following updates on top of the major update. The report didn't acknowledge this problem, which I think it should have, because how many bad major updates we have had over the last five years.
"Apple needs to fix these issues in order to keep the quality of their products up to the standards that they are well known for because they are traveling down a slippery slope that can open the door to going backwards in our progress that we have spent years of work on up to now. Also, the iOS Netflix app is very inconsistent with accessible updates. I have been a Netflix user of the iOS Netflix app for a while and I have noticed some updates that are fully accessible, but others are not usable at all because they are not accessible. I more often get the updates that are not accessible.
As a result, I find myself spending months at a time unable to use Netflix because of their inaccessible updates. I wish they were more consistent with their app updates so that those who use the app can have a much more normal and enjoyable experience."
Thank you for your email, Kendra. I think perceptions do vary widely, which is why the National Federation of the Blind passed a resolution last year on the concept of a defect equity framework so that we can go to these manufacturers and say, "If this bug persists, it would be the equivalent of this happening to a sighted person," and we have seen bugs in various manufacturers' operating systems over the years where the equivalent would be somebody's screen not working, and no software would survive an hour if it was pushed out and it broke the screen access for sighted people. My perception of the history is a little different from yours.
I remember going back to iOS 7 and 8 and operating systems like that, and there was some very serious bugs. I believe it was iOS 8 or it may have been 9, where there was that critical bug where some blind people couldn't even answer a phone call. You can imagine if you're in a professional setting and you can't answer a call, that's a debilitating bug.
The concept of a defect equity framework is really designed to be helpful to these manufacturers to help them prioritize those bugs that are of such a showstopping nature that they need to be addressed immediately. They are mission critical bugs. I don't recall anything that show-stopping coming out in the more recent major updates to iOS. That doesn't at all suggest that there aren't some rough edges. There definitely have been, but I can't recall anything quite that bad. We did go through a period where Braille was breaking regularly.
That does also seem to have stabilized a bit, but again, if there's something that you're trying to do with the phone and there is a bug, then the history doesn't really matter. If you can't do what you need to do and it's a critical function, that is really serious and we need to keep working with the developers to encourage them to address those bugs in a timely manner. Appreciate your thoughts. As we like to do, we're going to conclude with our tech tip here on Access On. This one comes in the form of an answer to a question from a listener.
Michael Panteli...: Hello, Jonathan. First of all, I'd like to say that in this space of podcasting, there are many people that do very good podcasts. There's only difference. You do them much, much, much better. Well done. I would just like to ask for some help. In 2021, you kindly helped me set up iCloud on Windows so I could find it using the TreeView navigation. I was wondering if it's possible to tell me how to do that again as I'm in the market to buy a new laptop, please. Keep up the wonderful work you do for us. Thank you so, so much and I hope Bonnie is doing.
Jonathan: Thank you very much, Michael, for the kind words. That's Michael Pantelides coming in from Australia and he is recording on a Zoom H2 essential. They really have a plethora of Zoom recorders coming out now with voice guidance. They just keep adding to that list and that's very impressive to see. I think what you might find, Michael, is that now, if you install iCloud on that new computer and good luck with this, nice to have a new gadget, you should just find those things there in the TreeView.
If you open File Explorer and then you shift tab and you press I for iCloud Drive and maybe another I for iCloud photos, they should appear for you. However, if that's not the case or if you want to pin something else to Quick Access, here is what you do. First of all, you've got to find the folder. In the case of this one, the iCloud Drive folder is going to be in your user folder, so you may want to just go to the run dialog and type C:/users/ and press enter.
You'll find your users there, including your own username. Go in there and you will find iCloud Drive as well as other things that you may want to pin to your Quick Access, so if you find something you want to pin, all you have to do is bring up the context menu. Shift plus F10 is the safest way to do that these days since the application key appears to be going away on some newer computers and you will find pin to Quick Access as one of the options there.
Choose that and from then on, you'll be able to open File Explorer. You will press shift tab to get to your TreeView, and if you press I from there, then iCloud Drive should now come up. I do this for a number of things I use regularly, including my Mantis, which I have paired to my PC. I just have it in my Quick Access and it makes it very easy to find it and take action with it.
That concludes this episode of Access On: the Technology Podcast of the National Federation of the Blind. To send in a contribution for a future episode, email us, attach an audio clip or just write it down and send it to [email protected]. That's [email protected]. To keep up to date with Access On, follow us on Mastodon. [email protected].
That's [email protected] on Mastodon. To subscribe to an announcements only email list about upcoming episodes, send a blank message to [email protected]. That's [email protected].
To learn more about the National Federation of the Blind, visit our website, nfb.org, or phone us, 410-659-9314. That's 410-659-9314, and be sure to check out the Nation's Blind podcast right from where you heard this podcast.