Also interesting: after playing with the small model, people viscerally feel that GPT-2 could be used to generate fake news —old.reddit.com/r/MachineLearning…
Ilya Sutskever
 @ilyasut
20 Mar 2019
old.reddit.com/r/MachineLearning… --- r/ML gets the small GPT2 to write imaginary news. I didn't expect the small model to be that coherent.

Mar 20, 2019 · 6:46 PM UTC

3
3
33
Replying to @gdb
I bet that it is quite easy to classify such GPT-output as fake with just few sentences of input. I'm working on such case now.
Replying to @gdb
Can you ask it too write fake academic papers. It might stumble upon something big.
Replying to @gdb
I'm training the model with everything. It's really cool. I trained with the chat history I have with some friends, in portuguese, and it's not that bad. I thought that it wouldn't work at all in portuguese.
1
1