Top all-time · #5647
Delip Rao
Personal blog · deliprao.com/archivesaiprogramming
Delip Rao's personal blog at deliprao.com covers ai and programming. Since 2016, 14 different Hacker News readers have linked it 17 times, most often “Chances Are Your [Machine Learning] Models Are Racist, Sexist, or Both”. The last post was in 2024, but the archive is still widely read.
Delip Rao: AI researcher, author, and founder. Two decades in machine learning and NLP, currently at the University of Pennsylvania. Author of Natural Language Processing with PyTorch (O'Reilly).
Most-linked posts?
- Chances Are Your [Machine Learning] Models Are Racist, Sexist, or Both3 people · 4 mentions · first linked 2016-03-08
- Science, Practice, and Frustratingly Simple Ideas2 people · 2 mentions · first linked 2016-04-03
- A Billion Words and the Limits of Language Modeling2 people · 2 mentions · first linked 2016-09-25
What Hacker News says?
The point is that the bias is linear, and my model is linear, so the model fixes things. The example is synthetic (so we could know what the right answer is and check if we recover it) so of course I put everything in. In the linked article, I explicitly reference a real world case where the same linear model was used to discover that grades and test scores are biased in favor of blacks: http://ftp.iza.org/dp8733.pd…
Comment linking /archives/129
on “Deep-Fried Data”Submitted Is BackPropagation Necessary? as a story · 6 points
HN discussion- nl · 2016-08-30
Disallowing the use of race as a signal in machine learning That sounds like a great idea, but it isn't as simple as it appears. Most systems won't have "race" directly encoded as a feature, but that is insufficient. See slide 22 from [1], further in depth discussion discussion from [2] or just this feature (which many systems could automatically discover): Feature6578 = Loc=EastOakland && Income<10k I don't know wh…
Comment linking /archives/129
on “Mathematician Says Big Data Is Causing a ‘Silent Financial Crisis’” Submitted Teach Don’t Document as a story · 3 points
HN discussionHi there, I wrote the article. We also looked at data on things like number of forks and contributors on GitHub as well as massive rise in questions on StackOverflow (thanks to Delip Rao's great post here http://deliprao.com/archives/168 and invaluable ecosystem analysis from Francois Chollet). These other things were cut for purposes of length and because we felt a mainstream reader (which is more Bloomberg's audie…
Comment linking /archives/168
on “Google Sprints Ahead in AI Building Blocks, Leaving Rivals Wary”Submitted Science, Practice, and Frustratingly Simple Ideas as a story · 3 points
HN discussionFor starters, any machine learning system that relies on word embeddings learned from biased corpora to make predictions (or to make decisions!) will necessarily be biased in favor of certain groups of people and against others. This is not true. Here's an oversimplified example. Suppose your machine learning system wants to predict something, e.g. loan repayment probabilities. One input might be a written evaluatio…
Comment linking /archives/129
on “Language contains human biases, and so will machines trained on language corpora”Submitted The Twelve Truths of Machine Learning for the Real World as a story · 2 points
HN discussionI'm reading Delip's followup post[1] and it reminds me how much of ANN stuff is till pretty much alchemy. [1] http://deliprao.com/archives/137
Comment linking /archives/137
on “Deep Networks with Stochastic Depth”Submitted Design Patterns for Production NLP Systems as a story · 2 points
HN discussion- nl · 2016-04-04
I suspect this was posted because of Delip Rao's write-up[1] (which I suggest might be a better link). It's a nice - if somewhat controversial - summary. 40% speedup on DNN training with state-of-the-art results. [1] http://deliprao.com/archives/134
Comment linking /archives/134
on “Deep Networks with Stochastic Depth” Submitted A Billion Words and the Limits of Language Modeling as a story · 2 points
HN discussionBy the way , as mentioned in this a few more general recent articles about the state of NLP worth reading are: http://deliprao.com/archives/301 https://www.skynettoday.com/editorials/state_of_nmt
Comment linking /archives/301
on “Google Translate's “Sinister Religious Prophecies”, Demystified”Submitted Chances Are Your [Machine Learning] Models Are Racist, Sexist, or Both as a story · 2 points
HN discussionBy the way , as mentioned in this a few more general recent articles about the state of NLP worth reading are: http://deliprao.com/archives/301 https://www.skynettoday.com/editorials/state_of_nmt
Comment linking /archives/301
on “Google Translate's 'Sinister Religious Prophecies', Demystified”Submitted A Billion Words and the Limits of Language Modeling as a story · 1 points
HN discussionSubmitted Word Embedding as a Learning to Rank Problem as a story · 1 points
HN discussion
