Phoneme level outputs from the Whisper API

1. Fine-tuning API for text-davinci-003, GPT-3.5-turbo 2. Increase input tokens from 4k to 32k 3. Support for languages other than English, e.g. Korean

Also another thing: I understand that chatgpt had a cut off at 2021. But it would be amazing if chatgpt had a way to keep up with updated developer docs. As a web3 developer, I use chatgpt daily to help me build software and this is something that would be a game changer

1. Auto login 2. Easy way to check if the session is active instead of typing Hello/hi every session (related to 1) 3. Option to cancel what sent before the response starts 4. Save the previously selected model (I need to switch to legacy every new session 5. Why I pay 20$

Allow gpt-3.5-turbo users to provide the beginning of the assistant completion response. Turbo is brittle and often ignores prompts. One example is generating JSON responses or other formatted responses.

Documentation. Some things are intuitive. Once you've used chatgpt or the dalle tool online, it's easy to understand the api. Other stuff isn't intuitive. Embedding, for example, it took @mckaywrigley to explain in a really easy to understand way

1. More max tokens 2. Faster response It is great even now, but still I want more.

Make Whisper able to label who is who. Would make interview transcribing much more automatic.

Hi, thanks! I’m so new, used ChatGPT to start learning code w/ swift &python. want: -Usage charts for individual API keys and also playground usage -Swift examples for api implementations -allow simple davinci style completions with turbo-3.5 -in-house incubator/learning lab 🐸

I’m ending many of my prompts with “respond with only JSON” style data type specifications. It would be awesome if there was a simple way to instruct the API to respond with various data types.