The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Accessibility and Computer Vision - #425
Nov 5, 2020
Guest
Venkatesh Potluri
Guest
Sina Baram
Guest
Cynthia Bennett
Guest
Chancey Fleet
Guest
Meredith Ringo Morris
In this engaging discussion, Meredith Ringo Morris moderates a diverse panel featuring accessibility advocate Chancey Fleet, AI fairness researcher Cynthia Bennett, inclusive design expert Sina Baram, and PhD student Venkatesh Potluri. They delve into the transformative role of AI in enhancing image accessibility for the blind, exploring tools like Microsoft's Seeing AI app. The panel highlights the need for ethical practices, diverse perspectives, and the balance between automated and human-driven solutions to ensure meaningful user experiences. Expect insights on privacy challenges, bias in AI, and the future of inclusivity in tech.
AI Snips
Chapters
Transcript
Episode notes
BeSpecular for Subjective Descriptions
- Chancey Fleet uses BeSpecular, a human-in-the-loop image description service.
- Multiple answers create a subjective understanding, particularly useful for art or fashion.
Protest Access for Sensitive Content
- Chancey Fleet uses Protest Access, a volunteer group providing descriptions of protest multimedia.
- This offers access to sensitive content without burdening volunteers on platforms like Be My Eyes.
Nuance and Subjectivity in Descriptions
- Sina Baram emphasizes the importance of nuance in image descriptions, especially in art.
- Subjectivity plays a key role; a 23-year-old Black man and a 45-year-old white woman will perceive art differently.