FAULT TEXT: The Adversities in Automated Alternative Text with Henri Helvetica
summary
Web images number 3.2 billion uploads a day while 2.2 billion people have visual impairments, yet automated alt text fails: ImageNet scraped 15 million photos for training but platforms like Instagram and Chrome produce vague or wrong captions — e.g. Chrome labels Porsche 911gt3 and automated alt misses a pregnant woman. Studies report 61% of landing pages with missing alt and error rates up to 35% for dark-skinned females, so the fix is better datasets and writing great, contextual alt text.
AI-written summary · about
chapters (6)
related talks
48mText-To-Image Accessibility
01h 00mAI and alt text
55mBridging the Digital Divide: Empowering Accessibility Through AI Innovations
09mTen Reasons Your Application Isn't Accessible and What You Can Do About It
16mThe positive alt-itude! An inclusion tool for your accessibility
22mAccessibility for Everyone
49mLost in Translation
17mPositive alt-itude! An Inclusion Tool for Your Accessibility