Google CEO Sundar Pichai stated in a blog post on Monday that Bard conversational AI was going to be tested with the intention of making it more accessible “in the coming weeks.”
The Language Model for Dialogue Applications (LaMDA) technology serves as the foundation for Google’s Bard, which has been under development for a while.
Pichai stated that “Bard aspires to integrate the depth of the knowledge of the world with the strength, intelligence, and creativity of our massive language models.”
He went on to say that the software “draws on information from the web to produce new, high-quality responses,” suggesting that it would provide responses that were current, something ChatGPT is unable to achieve.
Before the advent of ChatGPT, which was made available in late November, Google had been hesitant to introduce its own language-based AI out of concern about the reputational risk of making a premature release.
Researchers have shown that the system can spew out false information or rubbish on a potentially huge scale utilizing language models similar to Bard or ChatGPT.
In November, Meta, the owner of Facebook, was compelled to remove the release of its own massive language model called Galactica after three days because users immediately began sharing its biased and inaccurate results on social media.
Pichai emphasized that Bard’s responses would “reach a high bar for quality, safety, and groundedness in real world knowledge.”
In order to save computer power and reach a wider audience, Bard would source its responses from a constrained version of its underlying language model, similar to ChatGPT.
Google also said that customers would soon see AI-powered improvements in its search engine, which is crucial for its impending battle with Microsoft.
According to Pichai, modern solutions would “distill complex facts and different opinions into easily digestible formats.”
Generic AI-enhanced search engines “will deliver structured responses to inquiries and no longer links,” CNRS researcher Thierry Poibeau told AFP.
However, Poibeau noted that bots like ChatGPT “can give erroneous responses, which is frustrating for a search engine.”