Google’s own upcoming AI chatbot draws from the power of its search engine
Google announced on Monday that it is launching an AI-powered chatbot it’s calling Bard “in the coming weeks.” While this might look like a response to ChatGPT—OpenAI’s AI-powered chatbot that has been getting a lot of attention since it launched late last year—the reality is that Google has been developing AI tools for more than six years. And although these tools have not been previously made available to the public, now, that might start to change.
In the blog post announcing Bard, Google and Alphabet CEO Sundar Pichai writes that Google has been developing an “experimental conversational AI service” powered by its Language Model for Dialogue Applications or LaMDA. (That’s the AI model that one Google engineer tried to claim was sentient last summer.) Bard aims to “combine the breadth of the world’s knowledge with the power, intelligence and creativity of [Google’s] large language models” by drawing from information around the web and presenting it in fresh, easy to understand ways.
Pichai gives a few examples for how Bard can be used, such as getting ideas to help plan a friend’s baby shower, comparing two Oscar nominated movies, or getting suggestions for what new discoveries by the James Webb Space Telescope to discuss with a 9-year-old.
While Bard is only available to “trusted testers” right now, it is due to roll out to the general public over the next few weeks. Google has used its lightweight model version of LaMDA, which requires less computing power to operate, to allow it to serve more users, and thus get more feedback. Here at PopSci, we will jump in and try it out as soon as we get the chance.
Of course, Google’s end-goal is to use AI to improve its most important product: its search engine. In the blog post, Pichai highlights some of the AI tools it’s already using—including BERT and MUM—that help it understand the intricacies of human language. During the COVID pandemic, MUM, for example, was able to categorize over 800 possible names for 17 different vaccines in 50 different languages so Google could provide the most important and accurate health information.
Crucially, Pichai says that the way people use Google search is changing. “When people think of Google, they often think of turning to us for quick factual answers, like ‘how many keys does a piano have?’ But increasingly, people are turning to Google for deeper insights and understanding—like, ‘is the piano or guitar easier to learn, and how much practice does each need?’”
He sees Google’s latest AI technologies, like LaMDA and PaLM, as an opportunity to “deepen our understanding of information and turn it into useful knowledge more efficiently.” When faced with more complex questions where there is no one right answer, it can pull in different sources of information and present them in a logical way. According to Pichai, we will soon see AI-powered features in search that “distill complex information and multiple perspectives into easy-to-digest formats, so you can quickly understand the big picture and learn more from the web.”
Once or twice in the blog post, you get a sense that Pichai is perhaps frustrated with OpenAI’s prominence. While never name checking OpenAI or ChatGPT directly, he links to Google’s Transformer research project, calling it “field-defining” and “the basis of many of the generative AI applications you’re starting to see today,” which is entirely true. The “T” in ChatGPT and GPT-3 stands for Transformer; both rely heavily on research published by Google’s AI teams. But despite its research successes, Google isn’t the company with the widely discussed AI chatbot today. Maybe Bard’s presence will change that.