Tarun Gogineni

84 posts
Opens profile photo
Tarun Gogineni
@TarunGogineni
machine learning researcher
@OpenAI
papers.nips.cc/paper/2020/has…Joined December 2020
558 Following
2,253 Followers

Tarun Gogineni’s posts

Tarun Gogineni
@TarunGogineni
in fact there are few areas in tech where you get clearer reasons to build something; but you can't rely on self interested researchers to not employ motivated reasoning around it every day that autonomous driving is delayed is lives lost err on the side of launching faster
Quote
Liza Dixon ⚡️
@lizadixon
Replying to @lizadixon
There is no good reason to rush the dev of driving automation. The inconvenient truth is that we don’t have the data we need to make claims ab the net effects of L2 systems. Do they have the potential to increase road safety? Yes. Do they have the potential to cause crashes? Yes.
Show more
Tarun Gogineni
@TarunGogineni
Replying to
@TarunGogineni
even if the efficacy of L2 is right on the borderline where effect sizes are unclear the real world training data is invaluable to create better autonomous systems and shorten the timeline to full autonomy
Tarun Gogineni
@TarunGogineni
he’s right from a talent allocation perspective from another perspective there’s quite a few problems researchers are not incentivized to explore
Quote
(((ل()(ل() 'yoav))))👾
@yoavgo
i want to get into ML research, what topic would you recommend? nothing. now it is not time to get into ML research. now its time to either observe what others are doing, or to build innovative applications using established techniques, or both.
Tarun Gogineni
@TarunGogineni
Replying to
@TarunGogineni
the author fails to mention that 1) of course, biological males are more likely to contract heart health issues than females, making the imbalanced training data representative of real world and 2) that data augmentation can improve performance regardless of imbalance
Tarun Gogineni
@TarunGogineni
common criticism of NLP transformers is that they’re blank slatist and don’t encode chompskian universal grammars but there are in fact papers that do this and don’t perform nearly as well as the naive transformer
arxiv.org
Improved Semantic Representations From Tree-Structured Long...
Because of their superior ability to preserve sequence information over time, Long Short-Term Memory (LSTM) networks, a type of recurrent neural network with a more complex computational unit,...