miga89

7 ms·

1. 1.▲
   by miga89 2y ago

   How do the organisers keep the private test set private? Does openAI hand them the model for testing? If they use a model API, then surely OpenAI has access to the private test set questions and can include it in the next round of training?
2. 2.▲
   by miga89 3y ago

   It seems like the best way of figuring out how strong a new model is, is to look at the benchmarks published by a 3rd competitor. Want to know how well the new Google model performs compared to GPT-4? Look at the Claude benchmark table.
3. 3.▲
   by miga89 3y ago

   When doing these rent vs buy comparisons people often neglect the opportunity cost of investing the down payment and any difference between rent and mortgage (if it exists) Rule of thumb according to [0] is that, assuming historical returns
4. 4.▲
   by miga89 3y ago

   "Chance favors the prepared mind" - Louis Pasteur "Beware the quiet man. For while others speak, he watched. And while others act, he plans. And when they finally rest...he strikes" - Anonymous "Anxiety is the dizz