Comments by "Christopher Bruns" (@user-td4pf6rr2t) on "Fireship" channel.

  1. 5:22 That was the most fun things ive ever done with JavaScript.
    1
  2. gpt-4 only beats the coding benchmarks last i remember and even though benchmark gpt-4 only does better because nlp due to bpe compared to bert sentencewise. So even though gpt-4 is beating gemini benchmarks - i am certain this is just an affect of natural language interpreter. So bpe aigns with how people talk compared to bert who preprocesses the human language before it generates tokens. So again to reiterate gemini performs less because it takes what we say and filters out the bast query from the corpus of words sent compared to bpe were similarity is a bias (' paste the same code in 2 times with a single block repeated a third time will be how make gpt-4 demonstrate my explaination.')
    1