A Call for Human-Centered AI
summary
Argues that AI must be designed at three scales — user community society levels — with concrete examples: AlphaFold and ChatGPT succeed technically, COMPAS biases persist in criminal sentencing, Tesla autonomy may increase traffic per a Texas Triangle study, radiology automation fails when trained on ideal datasets. Recommends augment not replace (radiologists who use AI outperform those who don't), embedded ethics in coursework, and industry wins like Walmart theft detection, Stitch Fix personalization, and Citrine’s reported 500% return using AI.
AI-written summary · about
chapters (10)
- 0:00Human-centered AI intro
- 2:38Three principles
- 5:09Why AI for good fails
- 7:41Designing AI at scale
- 10:10Future of human-centered AI
- 13:42AI business success stories
- 16:30AI ethics and hysteria
- 18:41Teaching AI responsibility
- 20:48Corporate AI ethics
- 22:12Humanity in 50 years
related talks
31mA more human-centered approach to AI
17mHuman-centered AI
25mHumanizing Technology: Ethics and Design in the Age of AI
23mOpening up the box: the complexity of human interaction with AI systems
23mHuman-Driven Intelligence: Where AI & Automation Fall Short
20mHuman above AI
54mEthical AI Development: Leveraging Human Centered Design for Inclusive and Trustworthy Solutions
37mFinding the Problem to Solve: Human-centered AI