> Our empirical investigations show the effectiveness of our methods, which allow us to estimate the embedding size of OpenAI’s gpt-3.5-turbo to be about 4,096.
Discussion of the other paper is at https://news.ycombinator.com/item?id=39675735 but DeepMind seems to care more about helping their peer keep secrets and did not release GPT-3.5’s embedding size. Also the other paper is mentioned in Section 9 “Simultaneous discovery” of this paper in case anyone wonders.
1 comment
[ 5.1 ms ] story [ 12.3 ms ] thread> Our empirical investigations show the effectiveness of our methods, which allow us to estimate the embedding size of OpenAI’s gpt-3.5-turbo to be about 4,096.
Discussion of the other paper is at https://news.ycombinator.com/item?id=39675735 but DeepMind seems to care more about helping their peer keep secrets and did not release GPT-3.5’s embedding size. Also the other paper is mentioned in Section 9 “Simultaneous discovery” of this paper in case anyone wonders.