AI for Accessibility and Beyond  ## Metadata - Author: [[GitHub]] - Full Title: AI for Accessibility and Beyond - Category: #articles - Document Note: --- tags: - #AI - #Accessibility - #StaffEngineerPromotion Author: Panelists (Andrew, Joy, Scott, Shen, Blanca as moderator) Source: YouTube Webinar Reference: [AI for Accessibility and Beyond](https://www.youtube.com/watch?v=S40-jPBH820), [How to Get Promoted Beyond Staff Engineer](https://www.youtube.com/watch?v=lkY6OjHzXP0) date created: 2025-08-22 18:55 --- # Title: AI, Accessibility, and the Path Beyond Staff Engineer ## Table of Contents: - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📖 References](#-references) - [🗣️ Paraphrase](#-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB (Diversity, Equity, Inclusion, Belonging)](#-deib-diversity-equity-inclusion-and-belonging) - [✅ Logical Validity](#-logical-validity) - [🔍 DARVO Analysis](#-darvo-analysis) - [🔍 SEO Evaluation](#-seo-evaluation) - [🎯 Conclusion](#-conclusion) --- ## 📝 Summary: Two sessions intersect here: one on AI’s role in accessibility and another on advancing to principal engineer. Both highlight how *bridging gaps*—technological or organizational—defines impact. AI can close the disability divide but risks amplifying bias, while engineers beyond staff must clarify fuzziness, delegate effectively, and align with business needs. Both call for proactive influence, thoughtful ethics, and an inclusive vision. --- ## 📖 Main Content: AI for accessibility emphasizes technology as a tool to dismantle barriers rather than defining disability as inherent. Case studies (Be My Eyes, Intel ACAT) show how AI tools empower independence. Yet, risks of bias, privacy, and misrepresentation persist. The staff engineer panel, by contrast, frames career growth as shifting from execution to influence. Engineers progress by building clarity from ambiguity, networking across silos, and aligning contributions with strategic business impact. Promotion committees prioritize communication, breadth, and visibility, not just technical delivery. ### 🗒️ Notes **AI for Accessibility** - [00:11](https://www.youtube.com/watch?v=S40-jPBH820&t=11): Disability divide affects 1.3B people globally; AI offers empowerment but raises bias/privacy concerns. - [03:02](https://www.youtube.com/watch?v=S40-jPBH820&t=182): Disability framed as environmental barriers, not individual traits. - [08:03](https://www.youtube.com/watch?v=S40-jPBH820&t=483): AI tools (Be My Eyes, Intel ACAT) enable communication and employment. - [19:10](https://www.youtube.com/watch?v=S40-jPBH820&t=1150): Generative AI poised to aid real-time conversation for disabled users. - [25:33](https://www.youtube.com/watch?v=S40-jPBH820&t=1533): AI-generated code risks inaccessibility due to bias and false confidence. - [37:56](https://www.youtube.com/watch?v=S40-jPBH820&t=2276): Brain-computer interfaces raise urgent regulatory/privacy issues. **Promotion Beyond Staff Engineer** - [05:34](https://www.youtube.com/watch?v=lkY6OjHzXP0&t=334): Skill shift—clarity from fuzziness, vision-setting. - [07:04](https://www.youtube.com/watch?v=lkY6OjHzXP0&t=424): Delegation reframed as turning vague problems into actionable work. - [09:12](https://www.youtube.com/watch?v=lkY6OjHzXP0&t=552): Building peer networks to sell vision. - [13:36](https://www.youtube.com/watch?v=lkY6OjHzXP0&t=816): Promotions not always desirable; roles shift significantly. - [19:15](https://www.youtube.com/watch?v=lkY6OjHzXP0&t=1155): Committees value breadth and influence, not only execution. - [24:52](https://www.youtube.com/watch?v=lkY6OjHzXP0&t=1492): Mentorship via “personal boards” more effective than a single mentor. - [32:15](https://www.youtube.com/watch?v=lkY6OjHzXP0&t=1935): Visibility to leadership crucial—leaders may not know you personally. - [37:55](https://www.youtube.com/watch?v=lkY6OjHzXP0&t=2275): Job changes can accelerate promotion but carry risks. --- ## 📖 References: - University of Washington research on generative AI for accessibility. - Intel ACAT toolkit. - LeadDev Staff+ series. --- ## 🗣️ Paraphrase: 1. *Original*: “Eliminating barriers leads to zero disabilities.” **Paraphrase**: When societal and technological barriers vanish, disability as a limiting construct disappears. 2. *Original*: “Don’t ask for permission.” **Paraphrase**: Growth demands initiative—waiting for directives guarantees stagnation. --- ## 📰 Newsletter: 🦾 **AI Meets Accessibility** + 🚀 **Engineer Promotions** Two big conversations shaping our future: 1. **AI for Accessibility** — From tools like *Be My Eyes* to Intel’s *ACAT*, AI is reshaping independence for millions. But beware: algorithmic bias and privacy threats still loom. [Watch the session](https://www.youtube.com/watch?v=S40-jPBH820). 2. **Beyond Staff Engineer** — A candid LeadDev panel stressed that career growth at higher levels isn’t about lines of code. It’s about clarifying chaos, influencing across teams, and making your impact visible to leadership. [Full webinar here](https://www.youtube.com/watch?v=lkY6OjHzXP0). 👥 Whether you’re scaling open source or your career ladder, the lesson is the same: *bridges matter more than walls.* --- ## ✅ Action Items: - Apply *accessibility-first* lens to WordPress contributor tools and docs. - Build “personal board of mentors” model in The WPCC for sustained contributor growth. - Frame OSPO leadership efforts at GoDaddy around *clarity in ambiguity* and *visibility to executives*. - Use AI accessibility insights to strengthen WordPress accessibility community efforts. --- ## 🧪 Hypotheses: - AI tools improve accessibility but amplify inequities if trained on biased data. - Principal promotions correlate more with organizational trust than technical depth. - Mentorship boards outperform single mentors in staff+ career growth. ### 📊 Methodology: - Accessibility: audit AI models for bias in accessibility outputs. - Promotion: analyze staff+ career paths across organizations, measuring visibility and cross-team impact. ### ❓ Questions: - How can AI in WordPress accessibility plugins avoid bias and false confidence? - Can mentorship boards be formalized in OSS communities like WordPress? ### 🔮 Future Research: - Explore AI-enhanced GUI design for blind developers. - Compare OSS promotion pathways with corporate frameworks. ### 💡 Implications: - AI regulation (privacy + fairness) is urgent for accessibility. - OSS projects need structured recognition frameworks beyond code. --- ## 🧠 Learning Styles Assessment: - **Visual**: AI talk lacks diagrams; engineer panel had no visual progression frameworks → improve with infographics. - **Aural**: Both videos support auditory learners. - **Verbal**: Strong storytelling, metaphors. - **Kinesthetic**: Could add interactive scenarios (role-play: delegating fuzzy tasks). - **Logical**: Both use cause/effect reasoning; could deepen with models/frameworks. - **Social**: Panels are inherently interpersonal. - **Solitary**: Add reflection prompts: “Do I *want* promotion?” or “Where do I encounter accessibility barriers?” - **Naturalistic**: None—AI analogy to ecosystems or networks could improve comprehension. --- ## 📊 Accessibility Assessment: - **AI session**: Transcript present but images lack alt-text. - **Engineer session**: Clear transcript, but dense language may overwhelm neurodiverse readers. --- ## 🌍 DEIB (Diversity, Equity, Inclusion, Belonging): - AI: Framed disability as environmental, not individual—progressive. - Panel: Gender diversity present, but limited global representation. - Improvements: More inclusive examples (non-Western contexts, neurodiverse experiences). --- ## ✅ Logical Validity: - **Strong Reasoning**: - AI session: causal reasoning (barriers → disability). - Promotion panel: inductive rea - Summary: AI for accessibility is a topic that aims to bridge the disability divide by leveraging technology. The use of AI in accessibility has the potential to empower people with disabilities and improve their quality of life. However, there are concerns about biases in the data used to train AI models and the deprioritization of accessibility and privacy in the rush to deliver AI applications. The document highlights projects and advancements in AI for accessibility, including the Be My Eyes app and the Assistive Context Aware Toolkit. It also emphasizes the importance of personalized and contextual language model support and the need for accessible methods for validation and more research in the field. - URL: https://www.youtube.com/watch?v=S40-jPBH820 Please welcome Ed Summers. There are 1.3 billion people with disabilities as a group. We experience lower outcomes across many aspects of life, including health, education and employment. The World Bank coined a term for this state of affairs. They call it the disability divide. They also identified the most promising way for us to bridge the disability divide, and that is technology. My name is that Summers. I'm a blind developer. I'm heard of accessibility here at GitHub and I've never been more excited about leveraging technology to empower people with disabilities to live the lives we want to live. I'm particularly excited about recent advance advances in generative A.I.. Yet I'm also concerned. I'm concerned that the data we use to train models includes biases against people with disabilities. In addition, I'm concerned that as we rush to deliver AI in applications and tools, we might be deprioritize some of the fundamentals like accessibility and privacy over the next 38 minutes. We'll explore this space, this intersection between A.I. and accessibility. You'll learn about projects that are making a positive impact on people's lives today. We'll also explore some of the shortcomings. These are areas that need our immediate attention. And along the way, you'll learn from three invited experts that are working at the cutting edge of this domain. But first, we need to dig deeper into this concept of disability. What is a disability? Well, there are many definitions that are used in many different contexts. But my Northstar definition comes from the preamble to the United Nations Convention on the Rights for Persons with Disabilities. And that document states, quote, Disability results from the interaction between persons with impairments and attitudinal or environmental barriers that hinders their full and effective participation in society on an equal basis with others. End quote. And from that definition, I've derived three principles that guide all of my work on accessibility. Principle one Everyone deserves the opportunity to participate. When people with disabilities participate, we all benefit. For example, could you imagine music without Beethoven or physics without Einstein or the 20th century without Franklin Delano Roosevelt? Principle two Disabilities are not inherent to people. The definition states that disabilities result from the interaction action between people with impairments and barriers, not that disabilities are inherent to people with impairments. As an example, if you told me as a blind person that I had to read the prompts on the teleprompter to the front of stage, I would be disabled. There's no way I can do that. But if you give them to me in Braille, I'm enabled. Principle three We all have the power to prevent disabilities, and that's particularly true for the people in this audience, because you collectively create the technologies that are used by all of humanity. And if you unknowingly introduce barriers into those technologies, then those barriers will undoubtedly create disabilities for some people. That leads us to the accessibility equation. The accessibility equation is a really simple way to remember these concepts and their relationship. And it states that disabilities equals people multiplied by barriers. Now, if you have a lot of people, we have a lot of barriers than we obviously are going to get a lot of disabilities. But what if what if barriers equal zero? Well, disabilities with equals zero. Let's tie all this get all this together with a personal example. When I lost my vision, I encountered barriers everywhere I turned everywhere I turned everything I wanted to do. I experienced a barrier. But over time, I learned the tools and the tricks and the technologies that are used by people who are blind in order to be productive. However, still to this day, on a regular basis, I encounter situations where there is no work around and access to visual information is required. And without it I face a barrier and I'm blocked. But you know what? My iPhone contains a camera. It contains a connection to the cloud. Or there's there's infinite compute that we can use around models. And wouldn't it be nice if I could snap a picture on my iPhone, send it to the cloud to be processed, and get back information about the visual environment in front of me at just the right time, in just the right way that I need. Well, luckily that app exists today and it's called BE My Eyes. Let's roll this video. Mm hmm. Many thanks to the entire team at Be My Eyes for such innovative technology. This next example went a little closer to home. It's about a developer named Anton, and it shows how Anton is realizing his full potential using A.I.. Let's roll. Anton. If you'd like to learn more about Anton and other developers with disabilities that are building on GitHub, check out the Read Me project at GitHub dot com slash read me. Okay. It's time to invite out our invited experts. They'll come out in sequence and when they wrap, I'll come back out and finish things up with a short conclusion. So first up, we'll have Darryl Adams from Intel. They're also going to share some amazing recent advances on brain computer interfaces. Then we'll have Jennifer Mankoff from the University of Washington. Jennifer and her team have been looking at generative A.I. with a rigorous approach of academia. And finally, you'll get to hear from a true visionary in the field of accessibility. Joe Devon With that, let's get started with Darryl Adams from Intel Thanks, Ed. And well, I just want to start by noting just how transformative be my AI really is. I'm visually impaired and I have a pretty difficult time with many visual tasks, including seeing the details of images on my phone. So I've been using recently, I've been using be my AI to describe my images for me, and the results are just remarkable. It's just like a series of like wow moments over and over and it just kind of feels like magic. And I also get this feeling of connected ness that I did not even know I was missing. Now I want to talk more about human connection and the importance of communication and the role that technology can play to help people with severe disabilities stay connected to the world around them until has created the Assistive Context Aware Toolkit or ACAT. And this is a opensource platform that enables people with motor neuron disease or M.D. or related and related conditions to communicate with others through keyboard simulation, word prediction and voice synthesis. Now ACAT is the result of years of collaboration between Intel Labs and Professor Stephen Hawking. At its Essence account is a switch scanning interface, which is an assistive technology that is used for augmentative and alternative communications or AAC. And the general idea is to be able to get access to all the kind of the primary computing functions through through a single switch. So again, supports a variety of input modalities that address a range of user constraints so we can trigger that digital switch by using like facial gestures detected by a camera or proximity sensors or even just off the shelf mechanical buttons and switches. iChat is also a modular and extensible platform where developers can add actuation methods and even modify and even replace the beat the base level language models. Now, Intel is has created a brand new version of ACAT that we plan to release later this quarter. And as MMD and maybe ALS in particular, as these disease conditions progress, the it is common that a patient will eventually move toward a locked in state. And when you're in a locked in state, that means that you have no ability or no voluntary muscle movement and you're unable to speak. So in order to enable people in a locked in state to commit to continue to communicate, we're including brain computer interface as an input modality in this new version of ACAT. So how is that going to work? We're using a low cost off the shelf EEG headset from open BCI, and then we we start with a short collaboration process where we create brain signal patterns called event related potentials or ERP is we then train a classifier to to detect the ERP waves, and then we use that classifier to predict the user's intent. And so we've been able to also increase typing efficiency and reduce latency by presenting the user with three different modes of communication. We've got a sentence mode. We have a canned phrase mode and we have a shorthand mode. So in sentence mode, so a user can use full sentences to communicate with others. And we're using a hybrid A.I. model that can both retrieve text and generate completions at the same time. Then with the canned phrase mode, we are generally, we use a semantic similarity model to retrieve the matches from the database. And this mode is we think is going to be really useful for those kind of simple messages where you might want to communicate a basic need to a caregiver for example. And then in our shorthand mode, this is an interesting thing. So we're basically or creating key terms that are using subject verb, object extraction. And we think that this mode is really interesting because it infers that you have some level of, of communication context already existing. So think about communication between a partner or a spouse and you kind of know what's going on already. But this mode just kind of gives us that additional boost that gives you a little bit of of context in connection. It's more natural. And we think the combination of all of these three modes creates a powerful and flexible solution that will allow people to have to basically be able to communicate across a broad variety of contexts. So I'd like to leave you with our CAT project, Guiding Principles. And the first is that we are ACAT is really designed to assist people who are locked in or those individuals who are experiencing advanced disease progression. The second is that we want to enable low cost sensing through robust processing and language model fusion. Think of this as we want to make that hardware as inexpensive as possible. We want to push the intelligence into the software and we want to keep that software free and open source. And then lastly, we want to improve communications through personalized and contextual language model support. I do actually want to add one additional sneak peak that is not on the slide, and that is that in our in an upcoming release, we also want to give a cat ears. We want the system to be able to listen to the conversation and provide responses or response suggestions in real time. Based on that context of that conversation, we feel that will be key to being able to reduce the amount of of of silent the silence gap in this type of conversation. It will make things much more fluid. So our our our goal is to enable people experiencing severe disability to stay connected to the world around them, using effective and efficient means of communication. So if you're interested in doing meaningful work that is, both can have immediate and tangible impacts on the human experience, please visit GitHub dot com slash intel slash ACAT to learn more. Now I'd like to welcome Jennifer Mankoff from the University of Washington. She's going to share with us the important research her team is doing at the intersection of A.I. and disability. Hi. I'm really pleased to be here. I represent the Center for Research and Education on accessible technology and experiences at the University of Washington, and I'm looking forward to talking to you today about our work on generative AI and accessibility. The question that I'm here to try to explore is how generative I can impact accessibility for people with disabilities who are using it or who are using the results of what it produces. And to explore this, my team conducted a three month auto ethnography, a sort of case study last spring. A group of seven people, five of whom had disabilities work together and independently to use GI to address two specific goals. One was to try to solve our own accessibility problems, and the second was to make sure that the documents, programs and other things that we created were accessible to others. We use a variety of tools for this, including GitHub, Copilot, Mid Journey ChatGPT and others all available off the shelf at the time. And we use them for everything from creating programs to creating documents to creating images and many other things. I'm going to tell a few stories from our case study. So first, one of the tasks that we used for was summarization and extraction, and this is useful for all sorts of things. It's already in use today to support things like documentation, generation, to summarize conversations and chat or to summarize documents. And for our team it was useful for a variety of disabilities. One I'll focus on is brain fog, which can occur with chronic illness and other things that affect cognition, so it can make it difficult to process certain kinds of things or even painful. And unfortunately, I wasn't a perfect solution for that. In particular, as one of our team members said, every once in a while the guy nailed a summary, but often gave me completely incorrect, ablest answers. And this was particularly insidious when it took, for example, a nuanced argument and replaced it with an able a trope meaning something that was harmful or just biased against people with disabilities. Because the only way to tell that it had done that would be to know the details of what was in the original source document. But the whole reason I was being used was because that source document was not accessible. Another example is visual imagery creation. Again, this can support a variety of disabilities, including brain fog, again, of somebody who has Fantasia, which is a difficulty visualizing something in the mind's eye or blindness. And this is useful for diagram creation, document creation and many other things. Let me read you the prompt that was used to generate this image, and I did not memorize that part. So I'm going to turn around because it's tiny and the firm monitors are represented, representative group of people with a variety of disabilities looking happy, but not at a party, including at least one blind person, one person wearing a hearing aid, one person using a cane and one person in a wheelchair illustration format. Now, you might immediately notice something about this image that doesn't match that prompt. Let's look a little closer. So a couple of problems with this image. First, the wheelchair has some very strange mechanics, less visible, but the wheelchair users arm is actually in a cyborg sort of fashion attached to the wheelchair arm. You can also see a prosthetic limb not attached to a body. In fact, there's a hand on it and the hand is attached to an arm that is also not attached to a body. And although this was the second iteration of our project guide doubled down, not only are there balloons, but there is confetti and a hand waving money in the air at the party. So party on, we did at least four different prompts and we weren't able to solve these problems. Lastly, I want to give an example in design and development. So you guys already being used in the sphere, as we know and in general, programing is pretty accessible. I'm going to focus here on the story of a person who's blind. So lots of goods there already. But when it comes to front end development and data science, which are two really hot fields for jobs, harder to include people who are blind with the tools that are available today. So the blind member of our team was exploring things like Could I help describe the layout of a Gucci or the icons in it? And the answer was yes. It did pretty well at this actually. And then he turned to data visualization and he told us tried to generate a data visualization to identify the right set of parameters to train the LDA model that he was working on. Michael said that the graph didn't make any sense at all. So he is GitHub Copilot for this. And Copilot did a great job of generating code that with a small amount of debugging would run, but there weren't tools in place for him to verify that the results of this were actually correct. Without turning to a colleague for help. So let's talk about what this might mean for the future of guy and work. First of all, I made a lot of errors. Pretty much every case that we came up with, there were some errors. This is not hugely surprising. And some of these errors where things that we've already heard a lot of talk about, like hallucinations. But in addition, I made errors that we hear less about in the dialog today. For example, it's Abel ism. And one thing I want to highlight with one more story was that it was also falsely confident. So we had a group member, for example, who wanted to generate a visualization that was accessible to someone with a color vision deficiency, and they asked Guy, Tell me how to do this properly. Not only was it confident, but it was correct about exactly how to do it. However, when they then asked for a visualization to be generated, it was not accessible and when they asked it to apply these principles to fix it, it actually got worse. So one of the key messages I think to take away from this is that validation and verification are often inaccessible, particularly if you can't access either the source document or the result of your use of GUI. And this leads to some danger. This, combined with the errors that we saw, leads to some dangers that I want to highlight before talking about potential remedies. So two key dangers are that we are seeing a rapid change in how software development is done. And we already are dealing with a world where software is in the majority, not accessible apps and websites, for example. But now we're seeing a situation where the default stuff that is produced by AI may be inaccessible by default. And worse, if somebody who has good intentions but not a lot of knowledge asks AI whether or not something is accessible or how to make it accessible, they may be convinced by the false confidence of the guy that it's doing a great job. So we face a potential future where accessibility already not great goes down even further. However, we also have an opportunity if we're able to address this, to create software, apps and websites and other things that are actually born accessible by default. Because in generating default code could introduce exactly the things that are often missing today, and that's the future we'd like to see accomplished. So how do we get there? There's four things that we'd like to see happen. First, we need better training data. Guy has been trained on today's software, which again is not accessible, so we need to find ways to take the training data that is accessible and really bring it into how we train GUI. And there are methods for taking data sets and using them to sort of improve performance on certain metrics. In order to do this, we also need better metadata. We need to know which things are accessible and that's often not labeled. Next, we need accessible methods for validation, and this is particularly important if we want to continue to include developers who are disabled in the work that we do so that they are their perspective is present in what's happening. And then lastly, we need more studies. We need to understand better how we can address accessibility when using AI, what are the right prompts, what are the right ways to train it, what are the right ways to validate it, and so on. And we need to put those out there and guidance in a form that is easily understood and used by developers in the field today. So thank you very much for your time and I'm going to pass the baton on to Joe Devon, who's going to be speaking next. Thank you, Jennifer. Now imagine that you are on the beach and you're approaching the water. You pick up an object from the water and then suddenly it gets snatched away from you. Did anybody get a really strong visual imagery of that? Raise your hand. All right. We got a good, good chunk of people and now we're going to do a little test. I'm going to ask you three questions. Number one, what beach did you have in your mind? Number two is what object was pulled out of your hand. And number three is what pulled that object out of your hand. Who had those three ideas in their mind? When I asked that before I asked the question, raise your hand. I see only a couple of hands raised. And if you are among those people who raise their hands, you have a very good chance that you have hyper fantasia, which is a condition of very vivid imagery. And what's really interesting I've learned over time is that you think about vision physically, you could be blind. You could be through this whole spectrum from blind to being 20, 20, or even better. But in your mind's eye, a lot of these abilities that you have also have this spectrum, and you could be like, I am more on the Fantasia side where you are unable to create vivid imagery in your mind. And that is like blindness in the mind's eye. And now, because of studying all of these things and working on artificial intelligence this year, it's made me actually rethink what is the meaning of disability accessibility and its relationship to artificial intelligence. And when you think about it, what is artificial intelligence? It's actually trying to understand sensory input, which means visual recognition, automated speech recognition, that kind of thing. Trying to understand cognitive processing, just like Transformers can take, let's say, a scientific paper and simplify it for you. And then generative output, let's say mid journey or even robotics are starting to get mixed in with AI a lot more. So you can do things like motion and all of that is trying to imitate being just like a human being. Now think about what is disability. In part, it's kind of caused by a disruption in all of those three things. Sensory input, cognitive processing or thinking and generative output where you might have some mobility issues. So they're very well aligned and both sides can really learn from each other. But one of the big challenges that artificial intelligence has to overcome is this uncanny valley effect. And I just put up a slide of a humanoid robot that has a very wide grin that's kind of creepy and gives you the sense of that uncanny valley effect. And in order to get beyond that, artificial intelligence needs to understand the nuance of human beings and our abilities. And when you think about accessibility and professionals that have worked with people with disabilities, we've spent decades trying to understand all of the differences in our abilities. And, you know, in order to get that nuance right, who better to work with and study from than people with disabilities? And so I maintain that the the accessibility community, the researchers on the accessibility side and the researchers on the A.I. side, if both groups work together, we will be able to push things so much further from the accessibility side. I think we need more researchers who come in and say, maybe it's time for accessibility to not, oh, maybe we we sit there and we help artificial intelligence researchers to understand these different abilities a little bit better. And what's going to happen from the artificial intelligence side is we're going to create a lot more assistive technology that is far better than what we have now. So now I'm going to come back to another ability now that you understand the basics of why I brought up the the initial idea about the mind's eye. And you may have heard that some people do not have an inner monologue now. Now, my fiancee, for example, doesn't understand how I don't visualize things, but I don't understand how people don't have an inner monologue, because that's how I think before I'm speaking to you all, I say it in my head and then and then I speak. This is a condition called Anna, really, on which you can think of of as deafness in the mind's ear. Right. And there are many, many more examples that I don't have time to get into. But I've really realized that the accessible kitty community has little details that make a difference when you're doing artificial intelligence research. And when you think about even this this concept of thinking in your mind's eye, if you're not working with the deaf population, for example, you may not realize that some people think visually. And I have a wonderful example. So if you're deaf and you have never heard a word in your life, you might have sign language as your mother tongue. I have a friend who is a coda, which is a child of deaf adults, and he was always been able to hear. But his mother tongue is actually American Sign language because that's how he spoke to his parents and he thinks visually and when he does a report, he has to run it by his wife to translate it into the kind of thinking that most of us have. And so when you're developing models that are utilizing these abilities, it could be really helpful to work with people with disabilities and with our community in general. And now let's go to the other end of the spectrum and think about some MEG abilities is what is what I'm calling it. So there's Tetra chromosome. Most people have only three color cones and those mapped to the three primary colors and anybody that works with displays is familiar with RGV. Red, green and blue are the three primary colors. But what's actually quite interesting is that there is a minority of the population excuse me, that have four or primary colors, that have four color cones and therefore four primary colors. And what's absolutely amazing is they can see 100 million colors. And I would love in these speeches to put a picture, a rainbow picture on the screen with all of these colors in order to ask you, hey, can you see this color or how many colors do you see and do a test for your chroma? See, but what's super interesting is that even though we have retina displays and 4K and we're moving on to, you know, a billion colors and more that are invisible because every single screen is RGV based, you will not be able to present a rainbow on the screen that a Tetra Chroma can see. And I think it would be really I have more detail obviously on this, but actually anybody working on display technology, there can be some very interesting proof of concept in order to try and emulate being able to see this on a wider spectrum or maybe creating at least some displays for Tetra chromatic to look at. I'm sure that we're going to learn lots of things from it. And now in doing all of this work, I've spoken to different disability organizations and some of them have really gotten overwhelmed because there's a lot of AI researchers that are approaching them trying to get data. So for example, American sign, there is a ton of data, but there is a sign language for every language out there, and most of them there's not enough data. So what I'm trying to do is figure out how to do a public private partnership, create an open source platform with a legal framework around it, in order that when this data is gathered, first of all, people with disabilities are paid properly for their work there, there is certain protection. It should all be, in my opinion, open sourced. There should be some legal requirement that any that the data is used in a way that doesn't have bias against people with disabilities and that it's used for something good. And now you've heard Darryl mentioned brain computer interfaces and now we're we're really reaching a point where technology is starting to be able to read our thoughts. And this is something that we really have this this moment where we have to pay attention and draw a line in the sand because we've known that ad tech and social media has gone a little bit far in terms of of our privacy considerations. But now if your brain is going to be read, you know, our thoughts are really sacred. And I think that there should be some regulation around it in order to make sure that the data is private unless we agree to share it. Because the one thing that we don't want to see is that data is your your thoughts are read. The data gets sent over through the cloud and people can read your minds and then maybe manipulate you. So it kind of goes both ways. And I think that there should we all talking about regulation these days and a lot of the regulation that is being proposed is to close down the open source movement. And I think we need to take a minute and just say, well, are we doing the right thing here? Because when the pieces come and it's going to start with the accessibility community, but eventually it'll reach everybody, we should really make sure that that's open source, that it's that it's by law, open source, that it's audited and that all of us have the ability to audit the models and make sure that this line in the sand is not crossed. This is the time to do it. We can't wait any longer because we know what's going to happen and things are moving fast. And I'm going to end actually with Darryl was was up here and he gave me this line that I could never forget and I always end with it, which is that in the future, A.I. and technology, your your experience is going to be generated in real time, in a personalized private way in the format you need. And when you think about people who are blind, their life is very I audio based. And so you're going to see transformation into an audio format. If you're deaf, your life is very visual and you're going to see the transformers turn the experience into something visual. And then I think what comes after that is if you're deaf blind, you will get the communication via haptics. And all of this is where AI and accessibility together are taking the future. Thank you. Thank you. All right, give it up for Joe, Jennifer and Darryl. Thank you. There are 100 million developers building on GitHub. Collectively, you are building the technology that accelerates human progress. and when you do that, you're pushing back the boundaries of disability and expanding human ability. And I want to leave you. Thank you. I want to leave you with one request. When you build, please build without barriers. Thank you for your time, AI for Accessibility and Beyond, GitHub, 01hgtg2zg5tb100y1vxjrbeftq.