|
|
|
|
|
|
by omicron1
2398 days ago
|
|
|
From working with many ML researchers, it's not speculation. Some aren't very proud, but if you want to get ahead and stay competitive, you have to publish your un-replicable results. The results aren't usually completely fake, but they depend on enough dishonest sleight of hand that they might as well be. |
|
|