Google has introduced new capabilities for Bard that make it better at answering both logical and coding questions.
Although Microsoft’s Bing Chat seems to be the frontrunner when it comes to AI-powered search assistance, Google’s Bard is slowly catching up in terms of functionality and features. Mountain View recently enriched its AI search interface with image integration as well. Now, it’s updated Bard with better logical reasoning skills so you can answer math and coding questions more accurately.
For starters, if Bard answers a question in tabular format, the table can be exported to Google Sheets for your own analysis. In addition, Bard can also take advantage of a technology called “implicit code execution” where it can detect whether a prompt is computational and then generate code accordingly in the background to provide a more accurate response. Google says This makes it easier for Bard to answer questions like:
- What are the prime factors of 15,683,615?
- Calculate my savings growth rate
- Reverse “Lollipop” for me
Google says its new approach to solving mathematical and logical questions goes beyond the capabilities of traditional language large models (LLMs). She claims that traditional LLMs are smarter at text prediction and can display some impressive capabilities in this space, but they fall short when it comes to solving problems that require deep thinking and standard problem-solving techniques.
As such, Bard’s latest update incorporating implicit code execution capabilities essentially combines the functionality of both traditional LLMs and code-based computations to provide better answers overall. Writing code automatically in the background helped Bard improve accuracy by up to 30% when it came to solving calculation and math problems in Google’s internal datasets. However, Google has warned that Bard’s code may contain flaws, or it may not include the code in its response, so it’s important to use it at your own risk.
Google is clearly keen to make Bard the next big thing in the AI-assisted search space, cementing its position as a leader in web search. It recently migrated Bard to the PaLM 2 language model and expanded its availability to 180 countries as well.