OpenAI has announced the rollout of a new feature for ChatGPT that enables users to set a persistent personality and instruction set for the chatbot. This enhancement ensures that the chatbot consistently responds in a desired manner across all chat windows. Initially available as a beta feature to Plus subscribers in the US, OpenAI plans to make it more widely accessible in the coming months.

The ability to customize the AI’s personality is already available in the GPT-4 API, which powers ChatGPT. This functionality allows third-party developers to incorporate personality customization into their own applications and interfaces. For instance, an enterprise developer can build a search interface for local documents that strictly adopts formal tones and imposes restrictions on words and language to meet specific requirements.

OpenAI made this addition to ChatGPT in response to user feedback gathered during its recent global tour. One example provided by the company involved a teacher using ChatGPT to create lesson plans. Every time the teacher started a new chat, they had to repeat the statement “I’m a third-grade science teacher.” The new settings eliminate the need for such repetition and can also be used to share expertise in a particular field, avoiding unnecessary explanations in each new conversation with the chatbot.

Moreover, novel writers can utilize these settings to share character sheets or provide examples of previous writing in order to have the chatbot write in their own voice consistently. The spokesperson for OpenAI explained that ChatGPT will consider these custom instructions for every conversation going forward, eliminating the need to restate preferences or information in each interaction.

OpenAI has been consistently adding new features to ChatGPT. In March, the company introduced plug-ins that enable users to connect to external data sources, including flight information, research papers, and Wolfram Alpha. Furthermore, OpenAI enhanced the capabilities of ChatGPT by incorporating functions such as generating graphs and curating datasets. While the browsing feature powered by Bing was initially launched, it has since been withdrawn due to undisclosed issues.

OpenAI has also prioritized safety and privacy concerns in the development of ChatGPT. Measures have been implemented to address different ways users may instruct the model, ensuring compliance with usage policies. The model has the freedom to refuse or ignore instructions that could lead to policy violations. The instructions given to ChatGPT are used to improve the model’s performance and are also employed for future training. However, users have the option to disable this default feature using the data controls provided. OpenAI ensures that personal identifiers from custom instructions are removed before utilizing them to enhance model performance.

While ChatGPT offers undeniable benefits, there are potential risks that need to be considered. A survey conducted in February revealed that 70% of workers use ChatGPT at work without informing their employers. Additionally, a recent survey found that 58% of employees actively use ChatGPT to save time on daily tasks at work. Such usage may expose employers to legal and compliance issues, as cautioned by web intelligence company Oxylabs. Furthermore, privacy concerns arise due to the reuse of user inputs by AI services to enhance their systems, potentially leading to data leaks. The theft of user credentials could also pose risks if hackers gain access to the sensitive information stored in chat histories.

In the realm of AI, language model tools like ChatGPT are invaluable but still imperfect, capable of generating unsubstantiated claims and fabricating information sources. Therefore, it is crucial to be cautious with the utilization of these technologies, considering privacy and security implications.

Read more: Will AI save the UK Civil Service £100m over five years?

The post OpenAI Introduces Persistent Personality and Instruction Settings for ChatGPT appeared first on TS2 SPACE.