Join Barton Poulson for an in-depth discussion in this video Biased labeling of data, part of AI Accountability Essential Training.
- [Instructor] When you're trying to evaluate…the potential for bias in artificial intelligence,…there's a few things that you need to look at.…I want to start by mentioning the biased sampling of data.…Going back to the hotdog, not hotdog example…I brought up earlier, if you want to classify foods…then you need to include the entire range…of foods that you want to classify.…On the other hand, if the people creating the model…use only certain kinds of data,…then that's all it's able to learn.…Think about training phones and smart phones…to recognize the full range of accents in American English.…
You'd have to include a substantial amount of data…from every different kind of accent…and then that doesn't even introduce…the issue of bringing in French or Arabic…or Swedish or something else.…Also, it's important to remember that most algorithms…are very sensitive to imbalance in the data.…They like to have a 50/50 split…on the outcome category, but if it's a rare disease…you're trying to predict, then there are solutions.…
- Bias in AI
- Navigating the social challenges of AI
- Moral reasoning and relational ethics
- General Data Protection Regulation (GDPR) and AI
- Discrimination in data
- Liability and AI
- AI in life-and-death situations
- Confronting the challenges of AI as a developer, executive, and consumer
Skill Level Intermediate
1. The Context for AI
2. Technical Challenges of AI
3. Social Challenges of AI
4. Legal Challenges of AI
5. Safety Challenges of AI
AI in the military4m 36s
6. Confronting the Challenges of AI
Next steps2m 52s
- Mark as unwatched
- Mark all as unwatched
Are you sure you want to mark all the videos in this course as unwatched?
This will not affect your course history, your reports, or your certificates of completion for this course.Cancel
Take notes with your new membership!
Type in the entry box, then click Enter to save your note.
1:30Press on any video thumbnail to jump immediately to the timecode shown.
Notes are saved with you account but can also be exported as plain text, MS Word, PDF, Google Doc, or Evernote.