Is code-davinci-002 just the largest non-GPT-4 model in the GPT-4 scaling law experiment? | Manifold

Is code-davinci-002 just the largest non-GPT-4 model in the GPT-4 scaling law experiment?

Mini

3

Ṁ35

Feb 17

44%

chance

1D

1W

1M

ALL

https://arxiv.org/pdf/2303.08774.pdf

Resolves Yes/No when I have >90% confidence.

Get Ṁ1,000 play money

Sort by:

It's 4 orders of magnitude less compute than GPT-4! GPT-3 was 2 OOMs less compute.

Related questions

Will GPT-5 have over 1 trillion parameters?

Will GPT-4 be trained (roughly) compute-optimally using the best-known scaling laws at the time?

GPT-4 #5: Will GPT-4 be a dense model?

GPT-5 trained with >=24k GPUs?

How many parameters does GPT4o have?

OpenAI releases a model unquestionably named "GPT-4.5" or "GPT-4.5x" (where X varies) by mid 2025, hardcore legalism arc

-12% 1d85% chance

What will be true about GPT-5? (See description)

What hardware will GPT-5 be trained on?

What will be true about GPT-5?

Will the performance jump from GPT4->GPT5 be less than the one from GPT3->GPT4?

Related questions

Will GPT-5 have over 1 trillion parameters?

OpenAI releases a model unquestionably named "GPT-4.5" or "GPT-4.5x" (where X varies) by mid 2025, hardcore legalism arc

Will GPT-4 be trained (roughly) compute-optimally using the best-known scaling laws at the time?

What will be true about GPT-5? (See description)

GPT-4 #5: Will GPT-4 be a dense model?

What hardware will GPT-5 be trained on?

GPT-5 trained with >=24k GPUs?

What will be true about GPT-5?

How many parameters does GPT4o have?

Will the performance jump from GPT4->GPT5 be less than the one from GPT3->GPT4?