Google announced on Tuesday that Bard, its generative artificial intelligence, will have the ability to fact-check its answers and analyze users’ personal Google data as the tech giant scrambles to catch up to ChatGPT in popularity. The release last year of ChatGPT, a chatbot from Microsoft-backed (MSFT.O) OpenAI, sparked a race in the tech industry to give consumers access to generative AI technology that can be used to generate new information based on existing content. Google is putting in place some guardrails to prevent Bard from causing the kind of backlash that tarnished ChatGPT after it likened an Associated Press reporter to Hitler and tried to persuade him to divorce his wife. It’s also limiting how much interaction Bard can have with users to help keep it on track and not spin out of control.
The latest additions to the tool, which will be available via a new website that opens in a separate tab on smartphones, also include a feature that will let you upload photos and videos to Bard for it to use to create an image-based response. The company also allows users to ask for multiple drafts of a query response, which it says will improve its accuracy and clarity.
While these updates are good news for users of the free tool, Alphabet Inc’s (GOOGL.O) Google still trails far behind ChatGPT in terms of traffic, according to third-party data. In June, the company’s search engine accounted for 88 billion monthly visits, while the ChatGPT website had 1.8 billion visits in the same month.
That may explain why Google is working to beef up Bard’s abilities, especially since it will be integrated into the Google Search app on Pixel phones from early next year. It’s designed to help it compete with the burgeoning chatbot market while keeping its search engine the de facto gateway to the internet for billions of people and fueling digital ads that drive most of its profits.
As part of the update, Bard will also be able to cite sources for some of the information it pulls from the web in its responses. This is a welcome improvement over how the tool often misses the mark when answering questions with no clear answer, instead using a Knowledge Graph card that supplies a definition or overview of a person or place. It has plagued other generative AI tools, even ones based on the most advanced models, such as OpenAI’s GPT-4.
In addition to citing sources, Bard will be more explicit about the facts it offers. Previously, the tool sometimes cited websites such as Tom’s Guide and Phone Arena when answering a question but did not include links to those sites. Now, it will do so if it is “certain that the information is accurate,” the company said.
Bard is a powerful tool that can be used to plan a friend’s baby shower, compare Oscar-nominated movies, or get lunch ideas based on the ingredients in your fridge. It can also pique your curiosity by explaining new findings from NASA’s James Webb Space Telescope or help you drill on your soccer skills. But it’s also prone to hallucinating and faking information, which has also plagued other chatbots.