Thanks Maneesh for articulating some of the frustrations people are having with these systems! Completely agree that UI is one of the important features of doing generative AI well. The other two for me are data and verification. You can read more in my post here - https://embracingenigmas.substack.com/p/why-your-generative-ai-startup-will. Happy to collaborate if you are interested.
Thanks for this great article!
Like ChatGPT made the model a lot more powerful because it was usable, further study on interfaces with AI models will lead to larger impact of them.
Thanks Maneesh for articulating some of the frustrations people are having with these systems! Completely agree that UI is one of the important features of doing generative AI well. The other two for me are data and verification. You can read more in my post here - https://embracingenigmas.substack.com/p/why-your-generative-ai-startup-will. Happy to collaborate if you are interested.
If you are interested in further thoughts about interacting with generative AI (LLMs in particular) I highly recommend @geoffreylitt 's recent post:
https://www.geoffreylitt.com/2023/03/25/llm-end-user-programming.html
This just in: a randomly trained black box has a random black box-like interface. A local developer is utterly shocked.
DALLE-3 goes one big step closer to conversational interfaces but I still want GPT-4 to be able edit images.