@openai: You know how in certain books, the first word starts with a giant, stylized capital letter? As it turns out, those big, fancy letters can be a glitch-causing nightmare for screen-reading programs for the visually-impaired.   
Jaxsen Day, a visually-impaired PhD student at the University of Texas Austin, says the fancy-letter issue is just one example of the many problems faced by modern accessibility tools. 

Jaxsen used ChatGPT and Codex to solve a lot of those issues, building himself custom workflows that assist with everything from making academic research easier to access, to better interpreting funky, poorly-digitized PDFs.

Now, as part of his dissertation, he’s working with Mario Guerra, UT Austin’s Associate Vice President of AI Platforms and Innovation, to use AI to make reading materials more accessible campus-wide. 

And the pressure is on: under new federal rules, public universities face an April 2027 deadline to ensure their websites and digital materials—including many academic readings—meet WCAG 2.1 Level AA accessibility standards.

Achieving accessibility at a university is far more complex than it sounds. Nearly every department has its own process of storing and digitizing research materials, so accessibility tools need to be as adaptive as their users. 

But Jaxsen believes that if he can use ChatGPT and Codex to satisfy his own unique needs, there’s no reason whole departments can’t either. Now Jaxsen’s priority is sharing his AI-based methods campus-wide. 

And after Jaxsen’s dissertation is complete? While he’s not exactly sure what’s next, he’s at least confident that it’ll take a lot more than a big, fancy letter to slow him down.

Kendall McDowell :

Same company who said our ram inside our PCs is terrible to have.

requecw :

Reset sora 2 pls

𝓐𝓲𝓭3𝔁 :

so you made chatgpt?

ScReAm_Hleb :

PLS RESET SORA 2 DAY 1

Eva🩰🧸🫧 :

Did you create chatCPT?

Keeba :

˚‧꒰აᴀɴɢᴇʟ໒꒱ ‧˚ :

I love supporting beginner content creators 🥰

AZP :

yoo ask ceo if he wants to buy my ram price non negotiable 1 million each only

𝐈𝐒𝐇𝐅𝐀𝐐𝐔𝐄 𝐌𝐈𝐑𝐀𝐍𝐈 :

Chatgpt never die 😂❤️

pələng777 :

❤️❤️❤️❤️🥰🥰🥰🥰

Owlcore AI :

Respect

dimcost42 :

Never thought a fancy first letter could break a screen reader. Are AI models good enough now to read layouts like this correctly?

Keeba :

Keeba :

Keeba :

Keeba :

Keeba :

Keeba :

ChadolfGiggler :

Same company who said our ram inside our PCs is terrible to have. My man, if that’s what u think, all of ur data centers is also a waste of electricity and water.

Keeba :

Steeve Homex :

That’s great! Currently doing that as well.

toxicstrategies :

OK cool thanks bro. I wonder if you guys will ever answer my email or address the audio playback. I recommended over a year ago for the cognitive load when you’re trying to use text to speech and you’re trying to process information and you ask a question and your question is so complex the ChatGPT will literally spit out almost 20 links and 40 pages each worth of a PDF file for you to research and make your own conclusion from you know that 15 seconds back 15 seconds forward that we’re able to do on our phones that’s not available on desktop. or maybe even a pause function would help. for the real researchers, you know. And same with the side table on codex want to click on a link and it opens the side of your researching the resource the audio playback functional just completely disappears, but it’s all good guys. I know you got bigger problems to worry about.

acuarius23 :

primero

Aeris :

A before-and-after with footnotes and two-column pages would make the improvement really clear.

Ebbie Essa :

🔥🔥🔥M