Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

It also means our conversation is recorded by openai, doesn’t it?


This post reminds me of when I encountered people that were surprised to learn that the owner of a Telegram bot can read all messages sent to the bot.


Part of being an engineer is making products that act the way the users want them to act.

The fact that a Telegram bot owner can see everything the bot does is not what most users want/expect, and is therefore an engineering deficiency.

Obviously fixing this deficiency isn't easy...


If users want something functionally impossible then that’s an education thing not a product thing.

I fully expect that when I send text to a service that the provider can read all of it. Like of course they can, they have to process it.


A privacy protecting telegram bot isn't impossible. The code of the bot could be run locally on the users phone to generate the responses.

Or telegram themselves could run a "bot server", where bot owners could upload code to, but not inspect logs of messages going to/from each user.


>where bot owners could upload code to, but not inspect logs of messages going to/from each user.

"Because, trust us"

Honestly the moment a packet leaves your device privacy is gone.


> code of the bot could be run locally

How is that closer to a "telegram bot" and not just a "phone app"?


Yeah, PEBKAC bugs are the hardest to fix.


Not really, the chats could be stored locally in session data.


I’m confused. The AI does not process input in-browser.


If you use multiple devices they persist across all of them


Obviously but I’m saying a trade off could be made?


You're incongruously confused about this for your username.


Everyone misunderstood me, I couldn't be bothered fixing it.

It's fun to hallucinate anyway.


how do you think the sidebar works?


As a web developer, I'll be perfectly honest in expressing shock that anyone would assume it wasn't being recorded. Even in far less important spaces, basically every click and keystroke is sent to some analytics service, and if you block it on the client, there's generally a server-side gatherer to pass on what it can. If you ever submit anything to any modern website, it is a somewhat safe assumption that it is recorded somewhere.

This applies a little less if you're in a country affected by GDPR, but even then - request that data some time, it's so much more than some people realize.


> It also means our conversation is recorded by openai, doesn’t it?

That was already happening anyway. But it means something else too

- shared prompts will be scrapeable for open source

- shared prompts will have many continuations and social feedback around them - valuable for training? who knows


It saves all previous messages in an array of messages so it ‘remembers’ the conversation from the beginning. So when you post your message you’re also posting the complete message history. So it’s kind of not recording in the sense that it will give your conversation content to other users.


It is. The company I work for apparently have purchased a version of GPT4 that does not send data back to OpenAI. The conversation is still recorded by the folks who are maintaining it though.


>Data submitted through the OpenAI API is not used to train OpenAI models or improve OpenAI’s service offering. Data submitted through non-API consumer services ChatGPT or DALL·E may be used to improve our models.

https://openai.com/security

It's not a special version of OpenAI, it's the default behaviour when you pay for the API


If you trust them.


Just like you have to trust AWS/Azure and any SAAS service in existence.


I don't.


How about the people who made your keyboard? Do you trust them? If not, what do you do about it?


I speak every keystroke into OpenAI Whisper.


I make my own. Your move.


Is this the Azure deployment of GPT-4 or something else?


Yes, so far as I know.


If you thought it wasn't that would be hopelessly naive. Of course they record it. They will also read it, analyze it and do whatever they want with your conversations.


Your conversation history in a thread is the entire prompt for each new question you ask. When you log in, you can see all of your prior conversations. So, yes.


Chat history can be disabled in the settings menu, temporarily or permanently at least for Plus users, not sure about free users…


It disables it for you. But I would not bet on it also disabling it for OpenAI and even if it does today there is no guarantee it will be like that tomorrow.


It is deleted after 30 days, the same as their API


You know this how? Because they say so or because you've audited them?


As per this OpenAi blog post April 25, 2023.

'We've introduced the ability to turn off chat history in ChatGPT. Conversations that are started when chat history is disabled won’t be used to train and improve our models, and won’t appear in the history sidebar. These controls, which are rolling out to all users starting today, can be found in ChatGPT’s settings and can be changed at any time. We hope this provides an easier way to manage your data than our existing opt-out process. When chat history is disabled, we will retain new conversations for 30 days and review them only when needed to monitor for abuse, before permanently deleting.'

https://openai.com/blog/new-ways-to-manage-your-data-in-chat...


You have two options in the settings menu:

1. Share all your chat data with OpenAI

2. Disable saved chats entirely.

You want to have saved chats but don't agree with OpenAI also having access to them? Not possible.


You can also submit this Google Forms buried three links deep in their privacy policy: https://docs.google.com/forms/d/e/1FAIpQLScrnC-_A7JFs4LbIuze...


I use ChatBot-UI to achieve something like 1 and 2 together. It runs locally and uses the OpenAI API to provide a ChatGPT-like experience, but the chat history is saved locally in the browser. It also has a useful prompt management sidebar.

Of course, OpenAI saves the data regardless, but they say they don't use API requests for training data and they claim it's deleted after 30 days.

https://github.com/mckaywrigley/chatbot-ui

https://openai.com/policies/api-data-usage-policies


I've been waiting for GPT-4 API access for several weeks now. As soon as I get it I'll try something like this but who knows how long that will be.




Consider applying for YC's Winter 2026 batch! Applications are open till Nov 10

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: