News for hackers. Stuff that matters.

A comment on The LLama Effect: Leak Sparked a Series of Open Source Alternatives to ChatGPT. Parent comment

Re:The LLama Effect: Leak Sparked a Series of Open Source Al... by Closi on Sunday April 09, 2023 @04:27PM (#35506163)
I’ve got access to 4 and it’s a huge leap up from 3.5 - much more subtlety in the response, less hallucinations, less hitting a brick wall, but all of it adding up to a giant leap.
[ Parent | on HN ]
Re:The LLama Effect: Leak Sparked a Series of Open Source Al... by WXLCKNO on Sunday April 09, 2023 @05:08PM (#35506481)
It's funny how big the difference feels between 3.5 and 4 but also feels subtle. Like it's just doing what I expect it to versus seeing the limitations more clearly in 3.5.
[ Parent | on HN ]
Re:The LLama Effect: Leak Sparked a Series of Open Source Al... by novok on Sunday April 09, 2023 @10:24PM (#35508546)
It's also an order of magnitude larger in parameter count...
[ Parent | on HN ]
Re:The LLama Effect: Leak Sparked a Series of Open Source Al... by buryat on Sunday April 09, 2023 @10:53PM (#35508721)
unconfirmed, there’s a theory that it maybe similarly sized as gpt3.5 but with additional RLHF and because of other things it outperforms
[ Parent | on HN ]
Re:The LLama Effect: Leak Sparked a Series of Open Source Al... by Closi on Monday April 10, 2023 @05:53PM (#35517848)
Apparently the RLHF actually reduces it's capability, it just improves alignment/safety.
[ Parent | on HN ]
Re:The LLama Effect: Leak Sparked a Series of Open Source Al... by icelancer on Sunday April 09, 2023 @08:15PM (#35507746)
I agree. I think it depends on the context, though. I still use 3.5-turbo for text generation, content, ideas, etc, where it's completely suitable, fine, and even quite good.

However for any coding, complex analysis, or problems requiring calculation, there's no substitute for GPT-4. It blows 3 and 3.5 out of the water for code analysis, generation, debugging, and self-healing.

[ Parent | on HN ]