3 ms·
Discussion about the paper a while back: https://news.ycombinator.com/item?id=21084748 https://news.ycombinator.com/item?id=21084748. Conclusion seemed to be th
by augustt 6y ago
Discussion about the paper a while back: https://news.ycombinator.com/item?id=21084748 https://news.ycombinator.com/item?id=21084748. Conclusion seemed to be that the comparison against Mathematica was unfair since Mathematica's execution time was capped at 30s.
- hyperbovine 6y agoInteresting, since in my experience Mathematica's execution time is also lower-bounded by 30s.
- thomasahle 6y agoI don't think that was why it was considered unfair. The problem was rather that they tested on the same distribution the model had just seen 80 million examples from.