There Is No Spoon
summary
Generative AI produces plausible output but often gets practical details wrong: ChatGPT returns an “accessible” button with aria‑label misuse and a wrong type attribute, image description models improve over time and now enable real‑time and explorable descriptions on phones and wearables, third‑party accessibility overlays fail to add keyboard support for span‑as‑button elements and leak identifiable browser fingerprints, and one overlay vendor, Excessib, faces a $1,000,000 FTC fine for misleading claims.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:13Generative AI limitations
- 6:07Human role in AI
- 8:56AI hallucinations risks
- 11:01AI remediation dangers
- 17:41AI agents rise
- 20:53AI accessibility benefits
- 25:50AI image descriptions
- 28:16Real-time wearable AI
- 31:47AI data synthesis
- 32:53Future AI responsibility
related talks
31mThere Is No Spoon
32mAccessibility powered by AI
30mAI and Accessibility: the Good, the Bad, and the Bollocks
17mWhat AI Means for the Future of Accessibility and Disability Aides
47mAccessibility in the era of generative AI
34mAI and Accessibility: the Good, the Bad, and the Bollocks
48mAI and Disability: Progress or Exclusion?
57mCan Gen AI make the world Disability inclusive?