OpenAI gives developers more control over AI assistants
Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More
OpenAI quietly improved the controls in its Assistants API, making it easier for developers to work with files when building AI assistants.
The enhanced controls for File Search within the Assistants aim to help people building AI agents improve responses. The improvements let developers adjust how agents choose information from which to generate responses. They can inspect file search results and tune “the behavior of the file search tool’s ranker to change how relevant results must be before they are used to generate a response.”
OpenAI launched Assistants API in Nov. 2023, in what the company described as a baby step towards fully autonomous AI agents.
Assistants API allows developers to use OpenAI’s existing models with specific instructions when building assistants for different applications. It also allows them to use other tools within the OpenAI ecosystem. Assistants API also lets people create assistants that can “talk” with other agents.
While OpenAI told VentureBeat when the API first launched that Assistants API builds assistants that still need guidance — as opposed to agents that more else do tasks independently — it is a step in the right direction for developers looking to build more autonomous AI applications.
And it seems that many have been looking for a way to tune their assistants better as the response to OpenAI’s updates has been met with relief.
The agentic future
AI agents, where an AI can fulfill tasks automatically for users with very few instructions or prompting, are one of the big goals of many companies in the AI space. The idea is to take a lot of tedious tasks from humans, say booking a flight, and let the AI agent or assistant do it. For many enterprises, AI assistants could automatically fill up forms or take information from a dataset when they notice human employees starting specific activities such as talking to a customer.
Other companies also offer platforms to make it easier to build AI agents. Google recently made its software AI agent platform Oscar open source. Meanwhile, companies like Salesforce have begun releasing enterprise-specific agents for customers.
AI agents are still relatively new and still have tons of room for improvement, which is where features like OpenAI’s File Search controls come in. There is also no way yet to fully assess how accurate current AI agents are, and some reports have shown that benchmark tests on agents still lack some metrics.
Source link
