Google Enters The Generative AI Race With Bard, A Standalone Consumer Chatbot

Google launched its first standalone consumer AI product, a chatbot called Bard, on Tuesday (21 March).

It supports text-based questions and will operate independently from the Google search engine.

The launch of a new powerful language-based artificial intelligence (AI) by tech giants comes four months after OpenAI's ChatGPT public release triggered an industry rush to bring new AI technology to internet search.

OpenAI recently introduced GPT-4, a language model that can be accessed through ChatGPT's premium version or Microsoft's Bing search engine. Baidu also launched its chatbot, Ernie, positioning it as China's alternative to ChatGPT.

Productivity apps like Google's Workspace and Microsoft's Office 365 now integrate generative AI. The technology is also found in popular apps like Duolingo, opening the opportunity for millions of people to interact with it.

Bard by Google will exclusively offer English answers and grant access based on a wait-list in the US and the UK. It eliminates other computer codes or languages.

Google Research's vice-president, Zoubin Ghahramani, explained that they aim to receive feedback and increase Bard's user base gradually. This approach will allow them to test and learn before launching it widely.

Built on Google's AI technology, LaMDA, Bard is a language model for dialogue apps trained on web text data. With an added layer of training from Google Search results, it can provide reliable information with less chance of contradicting itself.

Chatbots using generative AI technology rely on vast amounts of human-generated text to provide believable responses to user queries, and have been among the initial consumer products developed.

Google has been lagging behind Microsoft in conversational AI development, with the latter making a significant investment in OpenAI in January.

This delay, according to some critics, can be attributed to Google's highly profitable search business, which prioritizes generating a single answer instead of introducing generative AI technologies.

Bard's lead, Jack Krawczyk, urged people to use the chatbot as an "experiment" for generating ideas and strategies, rather than as a search replacement.

He explained that GPT-4's limitations, such as hallucinations, made-up information and biases from its training data, were also present in the displayed model.

Models lack real-time web data and are inaccurate due to lagging behind the present.

Bard's responses lack citations, except when directly quoting from websites. However, users can confirm facts by using Bard's chatbot interface and clicking the "Google it" button, mentioned Krawczyk.