The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Language (Technology) Is Power: Exploring the Inherent Complexity of NLP Systems with Hal Daumé III - #395
Jul 27, 2020
Hal Daumé III, a professor at the University of Maryland and a principal researcher at Microsoft, dives into the intricate world of natural language processing. He shares his journey from mathematics to ethics in AI, stressing the importance of reducing bias in NLP. The discussion also highlights the need for inclusive data collection and effective evaluation frameworks. Daumé emphasizes that advancing fairness in AI requires collaboration and systemic change, calling for a commitment to sociolinguistic complexities in technology.
AI Snips
Chapters
Transcript
Episode notes
Accent and Identity
- Hal Daumé III was questioned about his accent during a talk.
- This experience highlighted how language can reveal social identities and create self-consciousness.
Dual Roles of Language
- Language has dual roles: constructing social identity and transmitting societal stereotypes.
- NLP systems must balance respecting identity expression while mitigating harmful stereotypes.
Language and Race
- Linguistic variations are linked to social constructs of race, but they aren't identical.
- Avoid assumptions about language use based on racial categories.