aub3bhat

16 ms·

91.
by aub3bhat 9y ago
Did you even read the article? """ This opinion piece by Mozilla Executive Chairwoman Mitchell Baker and Mozilla community member Ankit Gadgil """
92.
by aub3bhat 9y ago
You mean the Cincinnati riots of 2001 I can clearly see why USA should not have voter lists /s The riots happened in 2002 and NOT 2001 (at least get the year right), and unless you are arguing abolishing voter lists, or have a specific
93.
by aub3bhat 9y ago
>> They didn't rely on fingerprints, as they are easy to fake, without a high degree of technical sophistication Aadhar provides multiple levels of security, Number > Fingerprint > Iris >= OTP / Two-Factor. Finally
94.
by aub3bhat 9y ago
Unlike SSN number which lack associated biometric data, Aadhaar numbers lack utility without biometrics. A bunch of people read about SSN numbers "leaking" assume disclosure of Aadhar number is equally severe. Which is NOT AT ALL
95.
by aub3bhat 9y ago
All those myths are strawman arguments.
96.
by aub3bhat 9y ago
This is ridiculous fearmongering at a scale unseen before, given the enormous amount of corruption in Indian public distribution schemes, Aadhar has potential to be revolutionary in not only reducing corruption but also in providing poor pe
97.
by aub3bhat 9y ago
Is it traditional to announce startup shutdown on Friday (this & Sprig) just like layoffs?
98.
by aub3bhat 9y ago
Your argument is ridiculous. Conferences and journals are merely marketing venues, and there is no reason to slow down the field by clinging on to flawed review process. If you adopt this attitude you will only find co-authors and students
99.
by aub3bhat 9y ago
By generating the "code" are you merely inferring / representing UI layout as some kind of a tree. E.g. each button/panel/textbox gets detected, and an "RNN/LSTM" generate a tree structure using atten
100.
by aub3bhat 9y ago
Thanks for your answer. That clarifies my doubts regarding why OpenAI has not entered into more applied areas which can benefit from having a Non-Profit create tools / databases.
101.
by aub3bhat 9y ago
Can anyone at OpenAI explain the sole focus on RL while ignoring Vision, NLP ?
102.
by aub3bhat 9y ago
30 Rock did it first http://i.onionstatic.com/avclub/5224/22/16x9/1200.jpg
103.
by aub3bhat 9y ago
I don't deny that Kafka solves different problems and has excellent performance at high loads (100K msg/second). >> That doesn't necessarily mean people don't use Kafka for the wrong reason. Exactly what I meant.
104.
by aub3bhat 9y ago
I fail to understand the hype behind Kafka, my guess is that 99% of Kafka use cases can be handled by a simple RabbitMQ/Celery etc. But maybe its not considered cool enough?
105.
by aub3bhat 9y ago
Honestly this is an outdated article. In 2017 Hadoop is a more like OS/Platform/Ecosystem with FS (hdfs)/ Scheduler / Applications etc. Spark, Presto, Hive now ensure that you no longer have to write Map-Reduce jobs. Whi
106.
by aub3bhat 9y ago
Exactly my thoughts I am willing to accept the ridiculous pricing for web dynos considering that they provide a load-balancer / SSL / ability to quickly scale up etc. But running worker dynos or CI with same pricing just doesn&#x2
107.
by aub3bhat 9y ago
They are offering a dedicated 1000 TPU cluster for researchers, at least that's how I read it.
108.
by aub3bhat 9y ago
The cluster has 1000 TPUs
109.
by aub3bhat 9y ago
For comparison, at 180 teraflops per TPU (assumption: at fp32) this is equivalent to offering a cluster with 15,000 Titax Xp GPUs (1200$ per GPU = 18 Million $ worth of compute power). If TPU teraflops are reported at fp16, then the number
110.
by aub3bhat 9y ago
Funding is not knowing. There are always new markets which get ignored. Apple, Facebook, Google, Amazon, Microsoft did not compete with HP, Dell, Nokia they essentially entered a new market. Most ML startups don't offer a new product b
111.
by aub3bhat 9y ago
TESLA V100 is apparently ridiculously expensive at 65,0000$. In fact to an extent even NVidia has realized that there is more money in creating a GPU cloud from scratch rather than selling GPUs. I think the net losers are Apple, Amazon&#x2F
112.
by aub3bhat 9y ago
Frankly, I think your belief is due to lack of knowledge. As a current PhD student who knows several ML researchers at Google, there is fair bit of marketing involved in creating perception of Google supremacy. while no doubt Google is a le
113.
by aub3bhat 9y ago
180 teraflops per TPU seems great. For reference latest Titan X offers 12 TFLOPs [1] and upcoming AMD card for Deep Learning [2] offers 13 . Though its not clear if TPU performance is calculated at fp16 or fp32[2]. The best GPUs currentLY a
114.
by aub3bhat 9y ago
It's a great deal. Way more cheaper than some fly-by-night coding "academy" which charge 35K$ for 4 months. Also they offer a real degree, from a good US university accepted as a valid qualification by governments & corpo
115.
by aub3bhat 9y ago
I know they are different, but Google needs to realize their approach to consumer products does not transfers well to cloud computing.
116.
by aub3bhat 9y ago
I think Google just made job of some sales manager at AWS Rekognition way easier.
117.
by aub3bhat 9y ago
I really don't understand your arguments, let alone the overt negative tone here? >> 1. How will they maintain and extend that? Umm just like how everyone else does it. >> 2. H2O wrote its random forests and GBMs, neither o
118.
by aub3bhat 9y ago
I am using Celery workers which run a single tensorflow model in solo concurrency mode. The current workaround is to keep track of per process gpu fraction and then incorporate information about Total GPU memory available when assigning wo
119.
by aub3bhat 9y ago
I use per_process_gpu_memory_fraction extensively [1], but when you are writing code that runs on different GPUs (Titan vs 1080 vs 1070 vs K80) due to change in the total memory available you run risk of under allocating or over allocating
120.
by aub3bhat 9y ago
Is there any documentation regarding how this allocates / deallocates GPU Memory. While developing Deep Video Analytics [1] one of the biggest issue I have faced is TensorFlow refusing to deallocate GPU memory [2] (freeing up for other

More ›