Google has launched AI Test Kitchen, an application that allows users to test experimental AI-powered systems from the labs of the company before they make their way into production. Starting today, parties interested can fill out a sign-up form as AI Test Kitchen gradually becomes available to small groups in the US.
As announced at the I/O developer conference of Google earlier this year, AI Test Kitchen will serve rotating demos centered around new, cutting-edge AI technologies, all from within Google. The company stresses that these products aren’t finished products yet. However, they are intended to give a taste of Google’s innovations while offering the company an opportunity to study how they are used.
The first demos in AI Test Kitchen deal with the capabilities of the latest version of Google’s language model, Language Model for Dialogue Applications (LaMDA), that queries the web to respond to questions like a human. For example, one can name a place and have LaMDA offer routes or share a goal to ask LaMDA to break it down into a list of subtasks.
Google says it has added multiple layers of protection to AI Test Kitchen in order to minimize the risks like biases and toxic outputs around systems like LaMDA. As shown most recently by Meta’s BlenderBot 3.0, even the most advanced chatbots today can quickly go out of control, delving into offensive content and conspiracy theories when prompted with certainty text.
Google says that the systems within AI Test Kitchen will automatically detect and filter out objectionable content that might be sexually explicit, hateful, offensive, or divulge personal information. But the company warns that the offensive text might occasionally make it through.