Last Week in AI Mini Episode: Two Facial Recognition Stories, A Reckoning for NLP, and ”Self-Programming Computers
Aug 2, 2020
This week's highlights dive into the ethical dilemmas surrounding facial recognition tech, especially its use in marginalized communities. A study reveals that these algorithms struggle with face masks, raising concerns about accuracy. Meanwhile, NLP researchers are urged to rethink their goals, as current methods may be misguided. Excitement builds around a breakthrough in automated code generation, but it comes wrapped in clickbait. Dive into questions of justice and innovation in the AI landscape!
AI Snips
Chapters
Transcript
Episode notes
Rite Aid's Facial Recognition Controversy
- Rite Aid deployed facial recognition in 200 stores, focusing on lower-income, non-white areas.
- Tristan Jackson Stonkinos was misidentified as a shoplifter, highlighting racial bias concerns.
Masks Impede Facial Recognition
- Face masks significantly reduce the accuracy of facial recognition algorithms, by 5-50%.
- However, newer algorithms are being developed to address this challenge.
NLP's Reckoning
- Despite impressive benchmarks, natural language processing (NLP) systems are easily fooled.
- Jesse Dunietz argues that focusing on benchmarks distracts from true language comprehension.