8 ms·

Well, if that's the case, then the situation is worse than I imagined. I thought that training the model would be far more expensive computationally than using

Well, if that's the case, then the situation is worse than I imagined. I thought that training the model would be far more expensive computationally than using it to compute an 'answer'. Thanks.

MF-DOOM 2y ago
Training is much more computationally intensive than computing a _single_ answer. But not compared to running the model over time at scale for millions of users.
- MrMcCall 2y ago
  Thanks. That is precisely how I intuited the situation. 40+ years of programming FTW!