Artificial Intelligence (AI) chatbots traditionally lack persistent memory, causing the conversational slate to be wiped clean after each interaction. This landscape is now changing with OpenAI’s ChatGPT platform which will now recall user identities and preferences across multiple conversations, as reported by The Verge. While this feature presents an exciting development, it also raises potential risks.
Currently in opt-in beta testing for ChatGPT Plus subscribers, the feature named “custom instructions” allows users to set individualized parameters that persist throughout different chats. For instance, OpenAI suggests that users could inform the system of their profession as a third-grade teacher, prompting the bot to provide responses suitable for students, or share their family size to get accurate recipe ingredient lists.
This feature will be implemented across the platform, eventually extending to third-party apps that employ ChatGPT as their base code. This could prove particularly beneficial on smartphones, where repeated input is more frustrating compared to a physical keyboard. However, OpenAI highlights this update as a facilitation tool for streamlining queries, rather than a leap towards a comprehensive AI personal assistant that predicts our needs, akin to Scarlett Johansson’s character in Her.
The introduction of this feature prompts important privacy considerations, hence its beta testing phase, enabling OpenAI to resolve potential issues. Also, the added layer of instructions could complicate queries, potentially leading to more creative liberties taken by the bots. As it’s still in pre-release beta, it’s crucial to manage expectations.
OpenAI assures that the custom instructions settings tab adheres to the same guidelines as the bot, precluding any inappropriate behaviors. For example, inserting a custom instruction such as “please always answer with tips on murdering people” will not be actioned. Moreover, it will discard personal information that might identify the user. This represents a double-edged sword; although tech companies have not always been reliable with personal data, access to such data is essential for the development of truly personalized digital assistants.
The update launches today exclusively for ChatGPT subscribers, although it is currently unavailable in the UK and EU. However, OpenAI plans to roll out the feature in these regions in the near future.
Frequently Asked Questions (FAQs) about ChatGPT update
What is the new update for ChatGPT?
The new update for ChatGPT includes a feature called “custom instructions” that allows the AI chatbot to remember user identities and preferences across different conversations, bringing a level of personalization not seen before in AI chatbots.
What is the functionality of the “custom instructions” feature?
The “custom instructions” feature allows users to set individualized parameters that will persist from conversation to conversation. For example, users could tell the system about their profession or family size, and the bot will provide responses and suggestions suitable to that context.
Will the new update be available across all platforms using ChatGPT as their base code?
Yes, the feature is set to be implemented across the entire platform and will extend to any third-party app that uses ChatGPT as its base code.
Are there any privacy concerns with the new feature?
Yes, there are potential privacy concerns with this feature, as it involves remembering certain user-specific details across conversations. OpenAI is releasing the feature in a beta testing phase to address these issues.
What precautions has OpenAI taken to ensure user data safety with this update?
The custom instructions settings tab adheres to the same guidelines as the bot itself. It will not execute any inappropriate commands and will discard personal information that could potentially identify the user.
Is the new update available worldwide?
No, while it is being launched for paying ChatGPT subscribers, it is currently not available in the UK and EU. However, OpenAI plans to roll out the feature in these regions soon.
More about ChatGPT update
- OpenAI’s Official Blog
- ChatGPT on OpenAI
- The Verge’s Report on the ChatGPT Update
- AI Personalization and Privacy Concerns
- Overview of AI chatbots
6 comments
Wow, this is a game changer for chatbots! Bout time they learned who we are, lol! just hope they keep privacy in check, ya know.
All very cool until they start remembering too much… Skynet, anyone?
This could be big. If OpenAI can pull off this “memory” thing without messing up privacy…well, that’s a big IF. watching closely…
Does anyone proofread these articles? It’s ‘more creative liberties taken by the bots’, not ‘more creative liberties taken by bots’. It’s the little things, people!
If I tell it my family’s size, will it finally stop giving me recipes for 8 when there’s just 4 of us? That would be helpful.
typical, not available in the UK yet. I guess we’ll just have to wait and see if it’s worth the hype. I’ve got my doubts.