Chapter
Biases in Machine Learning Models
32:59 - 42:21 (09:21)
The speaker discusses the potential biases that arise in machine learning models, particularly with underrepresented minorities in training data, and the need for a democratically determined regulatory framework for its use, particularly in law enforcement.
Clips
The biases in machine learning models range from underrepresented minorities to a lack of democratic regulatory frameworks.
32:59 - 37:39 (04:39)
Summary
The biases in machine learning models range from underrepresented minorities to a lack of democratic regulatory frameworks. It is important to lean into these issues and work towards a reasonable solution.
ChapterBiases in Machine Learning Models
EpisodeKevin Scott: Microsoft CTO
PodcastLex Fridman Podcast
Implementing a verified chain of custody in social networks, using crypto and networks to have content signed, could provide a full chain of custody that accompanied every piece of content, allowing for better credibility of sources and the ability to detect deepfakes.
37:39 - 40:12 (02:33)
Summary
Implementing a verified chain of custody in social networks, using crypto and networks to have content signed, could provide a full chain of custody that accompanied every piece of content, allowing for better credibility of sources and the ability to detect deepfakes.
ChapterBiases in Machine Learning Models
EpisodeKevin Scott: Microsoft CTO
PodcastLex Fridman Podcast
Engineers are wired with a skeptical mindset, which is beneficial for the scientific community as they publish enough details in their experiments for other skeptics to try and reproduce the results.
40:12 - 42:21 (02:08)
Summary
Engineers are wired with a skeptical mindset, which is beneficial for the scientific community as they publish enough details in their experiments for other skeptics to try and reproduce the results.