Google Demos New Conversational AI Model and Opens AI Test Kitchen
Google unveiled the LaMDA 2 conversational AI model and introduced the new AI Test Kitchen app for experimenting with it at Google I/O 2022. LaMDA 2 (Language Model for Dialog Applications) improves on the version introduced at I/O last year with the help of thousands of users sharing feedback. The positive influence of that response prompted Google’s decision to launch the AI Test Kitchen as a hub for those who want to try out new AI models and tell the company about their experience.
LaMDA 2 extends Google’s work to design an AI model that can handle open-ended conversations with all of their unpredictability and illogical human meanderings. The model improves on the first LaMDA, itself a reinvention and upgrade of Meena, which Google called the most advanced chatbot ever at its 2020 debut. LaMDA 2 is trained on dialogue and taught to detect relationships between words, sentences, and longer passages so that it can better understand the context of the conversation and respond with reference to the whole conversation, not just the last thing the user said. LaMDA 2 offers a tool for designing conversational AI flexible enough to hop around topics without retraining.
“We’re continually working to advance our conversational capabilities. Conversation and natural language processing are powerful ways to make computers more accessible to everyone,” Google CEO Sundar Pichai said during his keynote address. “Last year, we introduced LaMDA, our generative language model for dialogue applications that can converse on any topic. Today, we are excited to announce LaMDA 2, our most advanced conversational AI yet.”
In The AI Kitchen
LaMDA 2 also benefitted from plenty of community testing and feedback for Google, so much so that the tech giant decided to essentially formalize the relationship. The AI Test Kitchen provides those who apply and Google chooses to test out new or unreleased AI tools and relay their idea and criticisms to Google directly.
“We are at the beginning of a journey to make models like these useful to people, and we feel a deep responsibility to get it right. To make progress, we need people to experience the technology and provide feedback. We opened LaMDA up to thousands of Googlers, who enjoyed testing it and seeing its capabilities. This yielded significant quality improvements, and led to a reduction in inaccurate or offensive responses,” Pichai said. “That’s why we’ve made AI Test Kitchen. It’s a new way to explore AI features with a broader audience. Inside the AI Test Kitchen, there are a few different experiences. Each is meant to give you a sense of what it might be like to have LaMDA in your hands and use it for things you care about.”
Google embedded three different kinds of LaMDA 2 interactions in the AI Test Kitchen for those interested in playing with the model. Talk About It is for conversing with the AI on a topic without getting diverted to other topics no matter what the user says. Though potentially workable for conversations on an infinite number of topics, the demo specifically focuses on dogs. The List It tool uses LaMDA to break down any goal or objective into a series of tasks that will help reach that goal, with the example of starting a garden generating a half dozen suggestions for a list including picking a style, planting seeds, watering, and getting rid of pests. Imagine It asks the AI to get creative in describing a concept shared by the user, a kind of short scene useful as a creative writing prompt.