0:54
OpenAI's code interpreter lets ChatGPT run code and test ideas in real time, filling knowledge gaps the model has. Community reverse-engineering suggests it uses a different model than GPT-4, which is why some call it GPT 4.5.
OpenAI's code interpreter lets ChatGPT run code and test ideas in real time, filling knowledge gaps the model has. Community reverse-engineering suggests it uses a different model than GPT-4, which is why some call it GPT 4.5.
ChatGPT 4.5 has been released and nobody noticed. But wait a minute, Igor, I haven't heard about this. Well, that's because it's not officially called GPT 4.5. Let me explain. OpenAI's code interpreter isn't just a new feature. It's a fundamental game changer. GPT3 brought pre training, GPT 3.5 added reinforcement, learning from human feedback. GPT4 introduced a mixture of expertise and now the code interpreter gave the AI a playground where it can learn and fill its own knowledge gaps. And this is crucial to understand. GPT could always produce code, but it couldn't run it. The code interpreter can, and that transforms ChatGPT from a passive assistant to an active participant. And it seems that the code interpreter accesses a completely different model from GPT4. How do we know this? Well, some clever people on the Internet have reverse engineered this and are 99% convinced that the model is not the same as GPT4. And that's why some people refer to this as GPT 4.5. Anyway, if you're curious to see a free beginner class on the code interpreter in action, check out the YouTube vid that I linked in the comments.
