HNHacker News
TopNewBestAskShowJobs

jonesn11

11 karma · joined July 2, 2019

submissionscomments
jonesn11··on DALL·E: Creating Images from Text
Someone made a replication of the github and it can be found here: https://github.com/lucidrains/DALLE-pytorch
jonesn11··on DALL·E: Creating Images from Text
This is a spot on point. My prediction is that it wouldn't be able to. Given its difficulty to generate correct counts of glasses, it seems as though it still struggles with systematic generalization and compositionality. As a point of reference, cherrypicking aside, it could model obscure but probably well-defined baby daikon radish in tutu walking dog, but couldn't model red on green on blue cubes. Maybe more sequential perception, action, video data or system-2 like paradigm, but it remains to be seen.
jonesn11··on When to assume neural networks can solve a problem
Not good on "small datasets", well in theory, but in practice transfer learning given a similar training distribution can take you quite far.
jonesn11··on When to assume neural networks can solve a problem
Here is a take from Google AI on when to use ML: https://developers.google.com/machine-learning/problem-frami...

"You should understand the problem well enough to list some potential solutions to test first to find the best model" and "it shouldn't be used for a problem that conventional programming can solve" like hard coding a symbolic rule.

← PreviousPage 2 of 2