Bard: Google brings two improvements to its AI chatbot Bard: What it means for users – Times of India

At I/O 2023, Google announced that it is now moving on to the next step and training its AI chatbot Bard on more powerful large language models (LLMs). The company has now announced two capabilities that have been upgraded as part of Bard’s improvement journey.
Bard gets better at logic and reasoning
The first capability is that Bard is getting better at mathematical tasks, coding questions and string manipulation. Google said that it has implemented a new technique called “implicit code execution” that helps the chatbot detect computational prompts and run code in the background.
As a result, Bard can now respond more accurately to mathematical tasks, coding questions and string manipulation prompts.

For example, if a user asks, “What are the prime factors of 15683615?” or “Calculate the growth rate of my savings”, Bard will be able to provide better answers.
Google claims that its new technique has improved the accuracy of Bard’s responses to computation-based word and math problems by approximately 30% in its internal challenge datasets.
Export action to Google Sheets
The second major update is the new export action to Google Sheets. For example, if users ask Bard to generate a table in its response, they can quickly transfer that data to Sheets.
Google already announced AI-powered tools in Sheets that can automate data classification and help users organise events and projects to track as well as analyse data faster.

Moreover, Bard can already export data generated in the chatbot to Gmail and Docs.
Google brings images to Bard
The latest development comes a few days after Google announced that Bard can now bring in images from Google Search. Users can ask Bard for images directly and the chatbot will show a source for each image.
Additionally, Bard also provides images in its response to offer better understanding. For example, if you ask Bard to “Show places to see in Lakshadweep,” it will show images along with text.