SUMMARY
On May 10, Google said that Bard would be rolled out to more than 180 countries across the globe, including India
Built on LaMDA, Bard uses Google's neural network architecture to identify patterns in a text to generate new output based on past experiences
Google also plans to soon integrate the AI chatbot with its suite of products such as Drive, Gmail, and Docs, among others
At the tech major’s flagship event Google I/O 2023 on May 10, key executives of the company congregated at California’s Shoreline Amphitheatre for the keynote address. Right after Google’s chief executive officer (CEO) Sundar Pichai, general manager for Google Assistant’s business unit Sissie Hsiao took the stage.
Half way through her speech, Hsiao made a big statement, which drew the applause of the audience. She announced that Bard would be rolled out to more than 180 countries across the globe, including India.
“Bard continues to rapidly improve and learn new abilities, and we want to let people around the world try it out and share their feedback. And so today, we are removing the wait list and opening up Bard to over 180 countries and territories,” Hsiao said.
Right afterwards, the Indian Twitterati went berserk, logging online to get their hands on Bard, which is touted as the single-biggest competitor of Microsoft-backed OpenAI’s ChatGPT. From questions about hot topics such as Assembly elections to creating code in Python, users online pitched a slew of questions to Bard, amusing some while disappointing others
In the making for years and unveiled two years earlier in 2021, the Language Model for Dialogue Applications (LaMDA) was launched publicly for select users in March this year under the name Bard. In simple words, Bard is Google’s answer to OpenAI’s ChatGPT.
What’s In Bard’s AI Heart?
Built in-house on LaMDA, Bard uses Google’s neural network architecture, called Transformers, and is designed around the premise of search. The network essentially identifies patterns in a large dataset of text and then iteratively keeps generating new output by learning from its past experience.
Trained on ‘natural-sounding’ conversational questions and responses, the AI chatbot has been envisaged with the idea of allowing more natural language queries for search, as against keywords previously.
This is achieved by deploying Transformers that comprise a series of hierarchical layers that process the input text one by one. Each layer has been designed to extract more and more information from a user’s query.
It begins by extracting information, such as syntax and structure of the text, and then progressively uses that data to learn statistical patterns and relationships between words and phrases in the query. It builds on its previous unsupervised training data to make choices and draw upon millions of lines of information it was earlier fed during the process of training.
In simple words, Bard deploys the LaMDA language model to break down a user’s question and understand its context. It then correlates the information with similar datasets on its LLM to understand the nuance of the query and then builds on the information drawn from the web to build a response. This answer is then streamlined in the format of a ‘conversational reply’ that a user is accustomed to and provides a better response than a regular search.
Trained on ‘natural-sounding’ conversational questions and responses, the AI chatbot has been envisaged with the idea of allowing more natural language queries for search, as against keywords previously.
But, while both ChatGPT and Google’s Bard have been trained on huge amounts of text, where the tech juggernaut gets the edge over the former seems to be the training data, which Google has collected and refined over the years.
Bard Has A Long Road Ahead
Just like ChatGPT, Google’s new AI chatbot could see deployment across the length and breadth of human spheres. Be it content generation or creative writing, Bard could help businesses and users save time and resources while building a consistent content stream.
It could also find usage in generating code for developers and help build engaging marketing campaigns with conversational slogans and taglines. Students can also use Bard for learning purposes, although the company claims that the AI chatbot is still far from perfect.
At its I/O 2023 event, Google also said that Bard would put more impetus on being ‘visual’ with its responses, meaning that the AI platform will also include images alongside text in its responses in future. The tech major is also looking at a full-scale integration of Bard with the search engine.
The tech major added that it will eventually also roll out features that will enable users to include images in their prompts, which will be achieved by integrating Google Lens with Bard. The AI chatbot will also be integrated with Google’s suite of products such as Drive, Gmail, and Docs, among other apps.
The India launch comes nearly three months after the generative artificial intelligence chatbot was unveiled in a disastrous public unveiling. In a live-stream that was marred by glitches and wrong answers, the aftermath turned out to be sour for parent Alphabet, which lost $100 Bn in market value as shares crashed as much as 9% during regular intraday trading the next day.
With competitor Microsoft-backed OpenAI way ahead of the curve currently, Google seems to have a lot of catching up to do. The delayed rollout comes amid a slew of challenges for Google in India, but, with technology and data by its side, the AI war seems to be heating up, even as new innovations in the domain continue to pile up.
Despite being nearly three months late to the party, Bard seems to be quite confident. When asked about its thoughts about ChatGPT, Google’s AI conversational platform called itself more powerful and versatile.
“Overall, Bard is a more powerful and versatile language model than ChatGPT. It has access to more data, it is built on a newer architecture, and it is capable of more tasks,” Bard answered, praising itself as the more superior AI conversational tool.
