GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
1–10 of 134 posts
Re: GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
#2$10 Million(M), not $10 Billion(B).
Re: GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
#3$10 Million(M), not $10 Billion(B).
Thank you, fixed! Also, direct link to paper: https://arxiv.org/pdf/2305.05862.pdf
Re: GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
#4Does anything currently beat GPT-4?
I saw some comments here say to check out Claude. From what I can tell, Claude hasn't figure out yet how to do the whole "generate Python code and run it in a Juptyer notebook" for math yet.
Re: GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
#5How do they know GPT-4 received no specialized financial training?
Re: GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
#6$10 Million(M), not $10 Billion(B).
Not looking forward for the times when an AI costs as much as an aircraft carrier.
Re: GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
#7How do they know GPT-4 received no specialized financial training?
Meaning, they used the same generalized foundation model that all of us have access to, with no special fine-tuning, no retrieval-augmented generation, etc.
Re: GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
#8Does anything currently beat GPT-4? I saw some comments here say to check out Claude. From what I can tell, Claude hasn't figure out yet how to do the whole "generate Python code and run it in a Juptyer notebook" for math yet.
Isn't that something you get from the infrastructure surrounding the llm? I thought the "running code" feature didn't need specific support from the llm, besides being able to output conforming json or code when asked to.
Re: GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
#9How do they know GPT-4 received no specialized financial training?
Meaning, they used the same generalized foundation model that all of us have access to, with no special fine-tuning, no retrieval-augmented generation, etc.
I don't understand your point. To me GPT-4 is not a foundation model, it's been highly tuned for the chat task. Nobody outside of OpenAI knows what that fine-tuning really involved. So it's impossible to say how much finance specific data it was trained on (in pre-training or fine-tuning) or whether finance specific tasks were involved in fine-tuning.
Re: GPT-4, without specialized training, beat a GPT-3.5 class model that cost $10M
#10Does anything currently beat GPT-4? I saw some comments here say to check out Claude. From what I can tell, Claude hasn't figure out yet how to do the whole "generate Python code and run it in a Juptyer notebook" for math yet.
A programming task where Mistral-large beats both GPT-4 and Claude Opus: https://sdk.vercel.ai/s/Thi7RDx9e8VOZo1Ee6We5 (only Mistral got the current syntax)
Although based on other tasks, overall, GPT-4 seems to be the best, but by a very small margin, so I cancelled my subscription. Although the native mobile app is really great.