Google is opening up its LaMDA conversational AI model to select US Android users.
LaMDA (Language Model for Dialogue Applications) debuted last year as a prototype AI system capable of deciphering a conversation's intent. It examines words in a sentence or paragraph to predict what will come next, generating long, open-ended conversations on potentially any topic.
The Test Kitchen app(Opens in a new window), introduced in May at Google's I/O developer conference alongside LaMDA 2, serves up a rotating set of experimental demos—or unfinished projects that give a taste of what's to come from the tech giant's artificial intelligence program.
Demos include "Imagine It," in which you name a place to generate "paths to explore your imagination"; "List It," where you can share a goal or topic for LaMDA to break down into a list of helpful subtasks; and "Talk About It (Dogs Edition)," which lets you explore the system's ability to stay on topic—in this case, dogs.
"We see a future where you can find the information you're looking for in the same conversational way you speak to friends and family," Googlers Tris Warkentin and Josh Woodward wrote in a blog post(Opens in a new window). "While there's still lots of work to be done before this type of human-computer interaction is possible, recent research breakthroughs in generative language models—inspired by the natural conversations of people—are accelerating our progress."
In a year of internal testing, Google uncovered several "harmful, yet subtle" flaws in the system. The model, for instance, may misunderstand the intent behind identity terms, failing to differentiate between benign and adversarial prompts. It also suffers from biases in its training data, generating
Read more on pcmag.com