Anita Kirkovska

Hey guys! As I'm working with a lot of customers on their AI workflows, I'm noticing two things: - GPT-4 is great for many tasks, but it is becoming very expensive in production - GPT-3.5 Turbo is a great and familiar alternative, but the results aren't as good So if you have a more complex LLM workflow, the balance here is to use GPT-4 for more complex tasks, and resort to GPT-3.5 Turbo...

Have you tried instructing Claude in the same way as you would GPT-4? Given the widespread use and familiarity with OpenAI's models, it's a common reflex. Yet, this approach doesn't quite hit the mark with Claude. Claude is trained with different methods/techniques, and should be instructed with specific instructions that cater to those differences. So, I looked into Anthropic's official...