With two recent advances, Google’s AI language model, Google Bard, is becoming stronger at logic and reasoning. The first enhancement is that Bard can now more effectively handle mathematical challenges, coding questions, and string manipulation. Additionally, Google Bard now allows you to export tables directly to Google Sheets.
To improve its reasoning and math abilities, Bard has included a new technique known as “implicit code execution.” Bard can recognise computational prompts and run programming in the background using this technique. As a result, Google believes that Bard can deliver more accurate answers to math-related questions, coding queries, and string manipulation requests. For example, it can now more successfully answer issues like discovering prime factors, estimating growth rates, and reversing words.
Why are there new enhancements?
The purpose of these enhancements is to provide Bard with improved reasoning and logic skills. While huge language models like Bard excel at language-related activities, they frequently struggle with more sophisticated problems that involve thinking and math skills. To address this, Bard now has the capacity to generate and execute code, combining quick and intuitive thinking.
Previously, Bard functioned solely on “System 1” thinking, making responses rapidly but without significant analysis. Bard can improve its response accuracy by introducing traditional code execution (“System 2”) into its approach.
Google states that using implicit code execution, Bard recognises prompts that can benefit from logical code, runs the code in the background, and uses the results to provide more correct responses. In internal tests, this combination technique improved the accuracy of Bard’s replies to computation-based problems by around 30%.