GPT-4

LLM that exhibits human-level performance

Instead of primarily text-based comprehension — we're now talking about any-to-any media transformation, on the fly. This is revolutionary:

Don't miss the developer demo livestream at 1pm PST today (March 14)!

I can't wait to see what will come out of GPT-4. I'm guessing automated AI product descriptions using image recognition. Better identification of situations based on photos (and videos?)

Very curious to know how it detects humor… when some humans can barely understand sarcasm 😏

Hey! I just wanted to reach out and say that GPT-4 sounds awesome! I noticed that its LSAT capabilities are much improved compared to GPT-3.5, and I'm excited to see how this can be leveraged for different applications.

That being said, I also noticed that GPT-4 might be less available to use due to cost, and I was wondering if there are any AI models in the works to help route to which model might make sense to use in which case.

Here are some questions that I have:

1. How can we optimize our prompts and questions to take advantage of GPT-4's LSAT capabilities, while also determining when it makes sense to use GPT-3.5 for tasks that require similar capabilities?

2. Are there any specific techniques or approaches that we should consider when designing prompts and questions for GPT-4, and how can we ensure that we are using the most appropriate model for each task?

3. Can you provide any guidance or best practices for switching between different models depending on the nature of the task, and how can we evaluate the effectiveness of each model?

4. Does the much larger context window of GPT-4 make it the clearest point for separating whether to use GPT-4 or not, such as when summarizing really long amounts of content?

Thank you so much for your hard work, and for building tools that are so empowering for people to be productive, efficient, and creative. I really appreciate hearing from OpenAI people such as talks online and learning more about the latest developments in AI.

I can't wait to get access to this. Well done, OpenAI team 🙌

How is the speed compared to GPT-3 models? I was surprised to see that gpt-3.5-turbo was so much faster than the vanilla gpt-3.5 model. Can we expect a similar speed to gpt-3.5-turbo?