Since unveiling its conversational AI Bard in February, Google has been working to improve the chatbot’s responses, after it provided incorrect information in its debut on Twitter.
More recently, we have seen the giant add generative AI functions to almost its entire suite of services, while access to the Bard chatbot has remained exclusive to a few. We have also seen some Pixel users receive invitations to test the Google bot, and now the company has said it is “beginning to open access to Bard.”
In a blog post that Google says “Bard helped us write,” product vice president Sissie Hsiao and research vice president Eli Collins invited people to sign up at bard.google.com. The company said it would begin implementing access for those in the US and UK starting yesterday and “expand over time to more countries and languages.”
Opening access to more people is “the next critical step in improving it,” the two said, noting that getting feedback from a larger test base is essential.
Like ChatGPT or Microsoft Bing AI, you’ll be able to talk to Bard as you would with a friend, using natural language instead of a series of keywords.
“You might ask Bard for tips on reaching your goal of reading more books this year, explain quantum physics in simple terms, or stimulate your creativity by highlighting a blog post,” Hsiao and Collins wrote.
You will soon be able to talk to Bard AI as well The two also said that “Bard is a direct interface to an LLM and we consider it a complementary experience to Google Search.” Based on the screenshots included in the announcement, Bard’s interface looks quite similar to Bing AI, with a few key differences.
At the bottom of each response, the Google version offers four buttons: thumbs up, thumbs down, a refresh arrow, and a “Google it” button. There is also an option in the upper right of the response that says “See more sketches.” Bing AI does not have these, but uses the space under each response as an area for source quoting.
While Bing AI runs on OpenAI’s GPT-4, Bard is powered by a “lightweight and optimized” version of Google’s LaMDA, and the company said that “it will be updated with newer and more capable models over time.” At its last I/O developer conference, Google unveiled an AI Test Kitchen app for people to interact with LaMDA 2.
The bottom input bar of the screen also has a few differences. Bard has a microphone symbol at the end, indicating that speech-to-text conversion could be accepted, while Bing does not. Meanwhile, Microsoft’s offering has an icon on the left side of the text input bar to clear the list when you want to start a new topic. Google does not have this.
It’s also worth noting that under Bard’s text field is a small line of letters that reads “Bard may display inaccurate or offensive information that does not reflect Google’s views.”
In its announcement, Google was careful to acknowledge that large language models (LLMs) like LaMDA are not perfect and that mistakes happen.
“For example, because they learn from a wide range of information that reflects biases and stereotypes in the real world, they sometimes show up in their results,” Hsiao and Collins wrote.
Google said it’s important to know about these challenges and noted that safety and quality are important aspects to consider.
“We’ve also built special features, such as limiting the number of turns in a dialog, to try to keep interactions useful and on-topic,” Hsiao and Collins wrote.
However, it’s unclear what the limit is on the number of turns in a dialog.
Despite all its potential limitations and possibilities for error, Google still believes Bard is worth the work, adding that it “will continue to improve Bard and add capabilities, including coding, more languages, and multimodal experiences.” Hsiao and Collins said that Bard’s help in writing the announcement post involved creating an outline and suggesting changes.