François Chollet on Twitter: "Machine learning has the potential to accelerate science. This paper from folks at Harvard and Princeton shows how deep learning can speed up progress towards fusion energy: https://t.co/FWTnpIO3KB It's great to see Keras and TensorFlow being used here."

Country	Code	For customers of
United States	40404	(any)
Canada	21212	(any)
United Kingdom	86444	Vodafone, Orange, 3, O2
Brazil	40404	Nextel, TIM
Haiti	40404	Digicel, Voila
Ireland	51210	Vodafone, O2
India	53000	Bharti Airtel, Videocon, Reliance
Indonesia	89887	AXIS, 3, Telkomsel, Indosat, XL Axiata
Italy	4880804	Wind
3424486444	Vodafone
» See SMS short codes for other countries

1. Variable Resistor‏ @ResistorQT 18 Apr 2019
  Replying to @fchollet
  
  You may literally save the world
  
  1 like
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Jari Safi‏ @safijari 18 Apr 2019
  Replying to @fchollet
  
  As a former nuclear engineer, I'd be thrilled if deep learning turns the ever constant "fusion is probably 30 years away" into a true statement.
  
  3 likes
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Tasha Harley‏ @TashaHarley86 18 Apr 2019
  Replying to @fchollet
  
  "prediction with long warning times" yes please
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Diogo R. Ferreira‏ @diogo4u 18 Apr 2019
  Replying to @fchollet
  
  Keras (for deep learning in nuclear fusion) was used here too:https://arxiv.org/abs/1802.02242
  
  6 likes
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Gaurav Ragtah‏ @gragtah 18 Apr 2019
  Replying to @fchollet
  
  Exciting times! Here's another paper on using CNNs to reconstruct the 2D plasma profile inside tokamaks + RNNs to predict plasma disruptions. From NVIDIA's GPU tech conference last yearhttps://www.profillic.com/paper/arxiv:1811.00333 …
  
  2 retweets 3 likes
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Mohamed ZAYED‏ @mzatwit 19 Apr 2019
  Replying to @fchollet
  
  Yeah, this is great. May be we can find some physical laws just by understanding what is done in deep learning models
  
  1 like
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Bobo‏ @notbyintent 19 Apr 2019
  Replying to @fchollet
  
  "stellarators are disruption resilient". Micro-instabilities have been around forever. Big doubts as to if tokamaks will ever work.
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Julian Kates-Harbeck‏ @juliankh 22 Apr 2019
  Replying to @fchollet
  
  It was a pleasure to work with Keras and Tensorflow on this project - greatly improved dev speed
  
  2 likes
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. David Pfau‏ @pfau 19 May 2019
  Replying to @fchollet
  
  It's probably impossible to stand athwart the deep learning hype, but it is worth noting that the results were basically a marginal improvement over the prior SotA. Nice work but not sure why it got into Nature.
  
  10 likes
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo