Reportedly, Bard, Google’s AI-powered chatbot, is gradually improving in functions involving logic and reasoning. Earlier in March, Google announced the launch of Bard, which stands as a competitor to ChatGPT. Just like OpenAI’s ChatGPT and Microsoft’s Bing, Bard is also a chatbot that is based on a large language model. One can give prompts to Bard and refine the answers further with follow-up questions.
Also Read: Google cracks down on office attendance, employees required to come at least three days a week
“You can use Bard to boost your productivity, accelerate your ideas, and fuel your curiosity. You might ask Bard to give you tips to reach your goal of reading more books this year, explain quantum physics in simple terms, or spark your creativity by outlining a blog post,” wrote Google VP of Product Sissie Hsiao and Google VP of Research Eli Collins in a blog post. When the world first gained access to Bard, all one could see was a long post written by the company’s CEO, Sundar Pichai. Bard is based on the tech giant’s own LaMDA, which is the Language Model for Dialogue Applications. Google is using a lightweight and optimised version of LaMDA.
According to a recent blog post, Bard is getting better at logic and reasoning. It is improving in performing mathematical tasks, coding questions, and string manipulation through a new technique known as “implicit code execution.” This upgrade comes with a new export action to Google Sheets. The blog further expands on the fact that large language models (LLMs) like Bard are primarily prediction engines. They generate responses based on predicting the words that are to come next. This feature makes them good email and essay writers. However, when it comes to software development, they might generate some errors.
AI models like ChatGPT and Bard were trained using an enormous amount of text from e-books, the web, and other resources. However, if we come to code-generating models like Copilot and CodeWhisperer, they were trained using exclusive code samples. In its attempt to address the coding and mathematics shortcomings in LLMs, Google has come up with “implicit code execution.” This allows Bard to write and execute its own code. It writes the code “under the hood,” tests it, and then uses the result to generate an accurate response.
Also Read: Getting AI innovation right requires policy framework ensuring accountability, trust: Google blog
On the basis of internal benchmarking, Google has asserted that the AI chatbot’s responses to computation-based word and mathematics problems have improved by 30% in comparison to the older version of Bard. However, when Bard was rolled out, it was a “disaster” that was accompanied by a Google ad showing a wrong answer by Bard. This led to an 8% drop in the company’s stock.
The update on Bard was tweeted by Sundar Pichai. The tweet mentioned the implicit code execution technique and Bard’s ability to run codes in the background when the bot detects computational prompts.
