News for hackers. Stuff that matters.
Re:GPT-4 by slt2021 on Tuesday March 14, 2023 @06:39PM (#35160142)
maybe we implement tokenizer+first layer in Javascript on client side and that is enough to preserve raw data on client side and send to GPT only first layer (which is a vector of float values anyway)

matrix gets decoded into text on the client side in Javascript, so we receive send and receive from chatGPT only vector of floats (obfuscation?)

Re:GPT-4 by geysersam on Tuesday March 14, 2023 @07:09PM (#35160683)
It's a good idea but it seems quite easy to invert the first layer mapping. And the output of the last layer you can easily steal just by doing whatever would've been done in the client.
Re:GPT-4 by nickpeterson on Tuesday March 14, 2023 @07:27PM (#35160993)
Could open ai just offer letting you upload a key and use it for interaction with the model? Basically encrypt the model with the key and all the request and responses are all secure?

I’m probably oversimplifying but it feels doable.

Re:GPT-4 by slt2021 on Tuesday March 14, 2023 @07:29PM (#35161040)
the goal is how to use chatGPT without sending plain text to OpenAI (to preserve privacy, make sure openai is unable to even see plain customer data)
Re:GPT-4 by andai on Tuesday March 14, 2023 @09:22PM (#35162393)
Maybe if we could speak with GPT-4 instead of OpenAI ;)
Re:GPT-4 by abudabi123 on Wednesday March 15, 2023 @03:26AM (#35165005)
Will the nonpareil paraquet make original discoveries and inventions from protein folding and stem cells results, GPT-X interfacing with DeepMind?