Two small requests with very large impact : 1) Replying to messages, similarly like in Discord "Reply" icon. 2) When ChatGPT not able to fully complete messages, a button or prompt "Finish last work" to handle it.

It should be able to go online

Please make it easy to export a chat and its history as one text document or markdown

Model fine-tuned for reasoning+summary+extraction rather than memorization. Example: I don't need a model to know the population of countries, as long as it can extract it from the input context that I give it.

Working on how LLMs can be used in Bayesian modeling and inference. It'd be great to have: - normalized log probabilities for when temperature ≠ 1.0 - stable log probabilities (currently differs across API requests, keeping text fixed)

Fine tuning for your latest models! ChatGPT Turbo is awesome -- just wish we were able to do more fine tuning using our data

More tokens per call 🙏 or support to parse larger texts in other ways

GPT4. Increased token limits. Especially for embeddings. Continue to drive the cost down. Faster responses. Confidence scores. First party vector database to store and serve embeddings. ChatGPT mobile. Increase the view width of ChatGPT in the browser.

A caching layer for embeddings and prompts - can reduce costs and maintenance on both sides

👉 Plan for the next decade: add reasoning ⇒get rid of hallucinations nitter.vloup.ch/VictorSenkevich/…