a bit scary though too
http://cs.stanford.edu/people/karpathy/deepimagesent/
This could be an incredible teaching tool. Imagine having children create their own captions and comparing them to the generated ones. The instructor could them guide the class through a discussion of which version is better and why.
Perhaps there is still hope for proper use of the language.
- Dale
Very impressive.
Wow, astounding. I wonder how it copes with more than one main subject/activity, although very impressive anyway.
Think more about a Google for pictures and movies, with the tags obtained by analyzing the image.
Or robots learning and interacting with the environment. Check this “kind of” experiment:
https://www.youtube.com/watch?v=1C-2dKe9NjI