dhairya

11 ms·

61.
Neural Magic Sues Facebook for Publishing Trade Secrets(americaninno.com)
2 points by dhairya 7y ago|0 comments
62.
Harvard Professor's Arrest Raises Questions About Scientific Openness(npr.org)
1 points by dhairya 7y ago|0 comments
63.
by dhairya 7y ago
It's intentional. Controversy is a strong signal for the youtube algorithm.
64.
by dhairya 7y ago
From the slides: Harms of AI for predicting social outcomes • Hunger for personal data • Massive transfer of power from domain experts & workers to unaccountable tech companies • Lack of explainability • Distracts from interventions • V
65.
by dhairya 7y ago
The oodaloop article is a summary of the NYtimes ( https://www.nytimes.com/2019/11/11/technology/artificial-int... ) article which is more detailed. Also pleasantly surprised in the NYTimes accurate descri
66.
by dhairya 7y ago
So a couple of disclaimers. I used to work at Harvard as an institutional research analyst (this was about 3ish years ago). My group worked closely with collecting and analyzing institutional data to advise senior leaders in strategic decis
67.
by dhairya 7y ago
It's worth noting that religion is not a demographic category that explicitly factors into admissions or collected. Students are not required to disclose their religion (sexuality for that matter as well). Those figures are collected t
68.
by dhairya 7y ago
If it's a shared resource why is there a need for one contribute in order to view the resource? The underlying implication of a private repo in this context is that the owner can at anytime lock the gate and make it pay to enter.
69.
by dhairya 7y ago
Just wanted to share this because it's one of the few resources that provides a great introduction to optimizing your models to scale and run faster. Each syllabus section has a great write up and set of exercises.
70.
NYU: Advanced Python for Data Science(nyu-cds.github.io)
4 points by dhairya 7y ago|1 comments
71.
by dhairya 7y ago
Deep neural nets have a natural tendency to overfit. Ideally, you have a held-out test which the model hasn't seen and only used after model has trained and tuned on the dev and validation sets. Often bad experimental models will repea
72.
by dhairya 7y ago
This is a great resource. Thank you for taking the time to write this and making it available to us!
73.
Nuts and Bolts of Applying Deep Learning (Andrew Ng, 2016)(youtube.com)
1 points by dhairya 7y ago|0 comments
74.
by dhairya 7y ago
I'm not sure I understand what this means. Most policy prescriptions rely on empirical research and hypothesis testing. Government is applied social science.
75.
by dhairya 7y ago
Happy to talk more offline, my email is my profile. The short answer is no because there are more complexities involved - both related to our specific use cases but really natural language in general. If that were the case NLP would be solv
76.
by dhairya 7y ago
as a side note, I love that you highlight regex in your new NLP course. There is an inherent tension between the probabilistic nature of models and the need for deterministic outputs in most production settings. Often if we can uncover ling
77.
by dhairya 7y ago
No worries, fair question. It worth noting that my job is not data analysis. So I do use data analysis to evaluate our metrics and model performance. Really none of it is really automatable. I'm working developing NLP features for our
78.
by dhairya 7y ago
So there is a difference between ML research and application. Being a practitioner doesn't require deep math knowledge that perhaps research would. Jeremy Howard's fastai course is a great example of how someone with a solid progr
79.
by dhairya 7y ago
Yes but the transition definitely doable and his advice is great. The key part of his advice is spending time experimenting, rapidly failing, and continuing to work on it with real world use case. Often the challenge is making the jump from
80.
VCS Excited About Natural Language Processing in 2019(slator.com)
2 points by dhairya 7y ago|0 comments
81.
by dhairya 7y ago
you add a linear classifier at the top to predict start and end positions of the answer span. The augmented model is trained on a qa dataset like squad to actually learn how to answer questions. hugging face has a simple implementation that
82.
Rotational Unit of Memory(super-ms.mit.edu)
1 points by dhairya 7y ago|0 comments
83.
by dhairya 8y ago
Also second the X1 extreme. I bought for my machine learning projects and work. It comes with a dinky GPU (1050 TI w/ 4GB) which is useful for running experimenting with small deep learning models locally. I was able to buy mine with 6
84.
by dhairya 8y ago
There was this guy who got an interview at Reddit using facebook targeting to target Reddit CEO Steve Huffman. http://twicsy-blog.tumblr.com/post/174063770074/how-i-target... >It turns out the Reddit CEO had a
85.
Owner rents $1,500-a-month Silicon Valley apartment for cats(nypost.com)
3 points by dhairya 8y ago|0 comments
86.
by dhairya 8y ago
would you mind linking to ones you are aware of. it would be super helpful.
87.
by dhairya 8y ago
Talla | Boston | Onsite | Multiple About Us Talla is working on creating knowledge-centric digital-coworkers for customer-facing teams. At Talla you will be working on hard problems for conversational interfaces (chat), rich text editing, a
88.
by dhairya 8y ago
Talla | Boston | Onsite | Multiple About Us Talla is working on creating knowledge-centric digital-coworkers for customer-facing teams. At Talla you will be working on hard problems for conversational interfaces (chat), rich text editing, a
89.
Google settled racism case from contractor 'treated like a terrorist'(theguardian.com)
4 points by dhairya 8y ago|0 comments
90.
by dhairya 8y ago
That didn't really come up. We were doing institutional research for a university to support senior leaders. Most of our work entailed working with transactional HR data, educational outcomes data, other information collected at the un

More ›