Comment by jpochtar

AI will solve web accessibility by screen readers that summarize visual content, ignoring ARIA and making it irrelevant. Multimodal GPT-4 can take a screenshot jpeg and answer questions about what’s in it (buttons, links, ads, headers, etc). The future of accessibility is rendering DOM to jpeg and asking GPT to be your eyes; we’ll look back on semantic markup as a failed idea that was never going to work