Home
> Uncategorized > My TED talk is live!
My TED talk is live!
August 22, 2017
It went up this morning, I hope you like it:
Categories: Uncategorized
mathbabe
Exploring and venting about quantitative issues
mathbabe 
Saw it and loved it! Thank you for making something so opaque suddenly clear.
LikeLiked by 3 people
Great, forwarded it to 15 friends.
LikeLiked by 2 people
Excellent!
LikeLiked by 2 people
This is great and so is your book! Thanks!
LikeLiked by 2 people
Loved it. I have your book in audio format, but I have to have it in dead tree form so I can re-read it.
LikeLiked by 2 people
Great work!! Posted link to talk. As you said people need to take political action to enforce transparency and accountability for those who create and implement these algorithms.
LikeLiked by 2 people
I read your book and loved it. Your talk is basically a 15 minute version of your book, but nevertheless I love it.
However, I am nervous about the “orchestra curtain” example you discussed. This needs to be handled with care. Omitting information like race or sex from a data set will not necessarily lead to a trained model that is color-blind or gender-neutral. These variables are likely to be correlated to other variables in the data set, and the algorithm may learn to use those other variables as a proxy for the missing variable, and thus learn to discriminate anyway (for example, the algorithm has no access to race data but since zip code is correlated with race that algorithm learns to racially discriminate by using zip codes).
A better approach would be to include variables such as sex or race when training a model so that you can remove their effects from other variables. But when the time comes to make predictions with the model, feed it false data; tell it that every applicant is white or male. This algorithm would in effect be less discriminatory than one that did not include this information when trained.
LikeLiked by 4 people
You’ve given an excellent reminder, ntGuardian. Simply omitting data doesn’t make algorithms any smarter. In fact, simply blaming algorithms misses the point that it’s better to optimize in new directions and in a clear eyed way. The training suggestion you mention, testing with false data, is more towards the benefit of a double-blind experiment, and less towards a color-blind (eyes closed head in the sand) approach.
LikeLiked by 1 person
I had a different take on the orchestra curtain example because she qualified it using the phrase ‘not distracted’ and ‘what’s important ‘
In pre-WW2 a German General is quoted as describing his generals ranks as having 2 traits
from the list clever, lazy, stupid, industrious
The clever lazy generals were said to be the best because they were not distracted and could focus on what was important
LikeLike
Outstanding presentation. This is important work you’re doing, especially so as I know of no one else doing it. Glad to see you doing it so well.
I hope everyone will “like” and share this talk; it’s relevant and important to every one of us.
LikeLiked by 2 people
Heard it this morning on the Ted Podcast. Enjoyed it and will share it along.
LikeLiked by 2 people
Ms. O’Neil: Speaking as a retired black male lawyer who is now studying Statistical Learning, Applied Probability theory and Machine Learning using Python…you are, and for a very long time have been, a stellar spirit-soul!!
LikeLiked by 3 people
This is an important and really terrific talk! After watching it today, I ordered a copy of your book. I’m looking forward to reading it. Thank you for the work that you’re doing.
LikeLiked by 2 people
Wonderful TED talk and so timely. I do a significant amount of data analysis as part of my job so I have an ‘above average’ appreciation for your message. In the natural world, mutation is central to evolution. Algorithms reject mutations, either as errors or outliers, so Big Data, in my opinion, will eventually lead to stagnation in so many ways. The hubris around ‘Big Data’ has been painful to observe.
LikeLiked by 3 people
Lovely talk. An eye opener. I hope those audits come true. Keep up the good work.
LikeLiked by 3 people
Thank you for a very informative talk. “Blind” being the operative word. Agree that big data need to be held accountable by transparency and by people being informed enough to demand understanding of the algorithm.
LikeLiked by 1 person
Reblogged this on Opportunities And Ideas For Home Business and commented:
After listening to Cathy O’Neil On Ted Talks the topic being “The Era Of Blind Faith In Big Data Must End,” It does give food for thought, and requires another look at how data is acquired and used. Looking at it from her perspective the use of data collected is not without bias, unintentionally or otherwise and needs to be looked at more closely for integrity checks.
LikeLiked by 2 people
Loved it, really did
LikeLiked by 1 person
Reblogged this on The Curious Scribbler and commented:
Algorithms reflect the human biases of their designers. This matters because algorithms may decide what kind of credit card you can apply for or whether you get a job interview. Thank you Cathy O’Neil for unpacking the black box.
LikeLiked by 1 person
This is such a relevant and important topic. Thanks for sharing your talk. Keep up the good work!
LikeLiked by 1 person
Yay!! Congratulations!!
LikeLike
I’m hearing impaired, do you have the text available?
LikeLike
If you watch it on the TED site, you can either watch the video with closed captions, or you can read the transcript. It’s worth a read!
LikeLiked by 1 person
Great talk. Congrats Cathy.
LikeLiked by 1 person