Andrew Carr (e/🤸) (@andrew_n_carr) 's Twitter Profile
Andrew Carr (e/🤸)

@andrew_n_carr

co-founder leading science @getcartwheel co-founder advisor @arcade_ai writer @tldrnewsletter Past: Codex @OpenAI, Brain @GoogleAI, world ranked Tetris player

ID: 3378986176

linkhttps://getcartwheel.com calendar_today16-07-2015 15:36:33

8,8K Tweet

21,21K Followers

4,4K Following

Andrew Carr (e/🤸) (@andrew_n_carr) 's Twitter Profile Photo
Andrew Carr (e/🤸)
@andrew_n_carr

2 years ago

Language models are bad a basic math. GPT-4 has right around 0% accuracy rate on 5 digit multiplication. Most open models can't even add. Why is that? There are a few reasons why numbers are hard. The main one is Tokenization. When training a tokenizer from scratch, you take

Language models are bad a basic math. GPT-4 has right around 0% accuracy rate on 5 digit multiplication. Most open models can't even add. Why is that? There are a few reasons why numbers are hard. The main one is Tokenization. When training a tokenizer from scratch, you take