Gone are the days when you had to guide your AI agents to browse and select APIs to use. Former Y Combinator and Stanford, David Yue, along with Finsam Samson, introduced Self-Learning Agent for Performing APIs (SLAPA), a model that automatically searches for API documentation, learns information, and creates API calls, in real-time.
SLAPA is trained to learn from itself. If it retrieves wrong information, it makes API calls again until it works. The information keeps getting stored in the model so users do not need to feed it information every time. As more users start using it and feeding in specific APIs, SLAPA will keep accumulating the information, while also becoming more comprehensive with time.
The ability for the model to know how to use tools is what makes it stand out. The creator of SLAPA explained in a Twitter thread that recently models like Toolformer have demonstrated that language models can be taught when they need to use the tools, but SLAPA takes it several steps further by teaching the model how to use the APIs.
Apart from the love from the community on Hacker News, a lot of people have expressed concerns about this technology. Some are even comparing it with Skynet, claiming that “AI is getting scary now” and making them anxious.