Google Bard: How ChatGPT's rival has improved since initial blunder
Google unveiled Bard, its answer to OpenAI's sensational chatbot ChatGPT, in February this year. Anticipation made way for disappointment as Bard's promotional video showed the chatbot answering a question incorrectly. The failed demo cost Google's parent Alphabet $100 billion in valuation and amplified the noise about Google's inability to challenge the OpenAI-Microsoft tandem. But Bard has improved since then. Let's see how.
Why does this story matter?
For years, Google was the leading force in artificial intelligence (AI). However, the company was hesitant to release a user-facing application to showcase its advances in AI. Google's hesitancy became an opportunity for OpenAI. ChatGPT's overnight success made Google squirm. This resulted in a rushed release of Bard, which had the opposite effect on the tech giant's credibility.
Failed demo prompted the company to start internal testing
The failed demo prompted Google to slow down with Bard. Later that month, the company started the internal testing of the chatbot. Google CEO Sundar Pichai asked employees to spend two to four hours on improving Bard. The company then sent a detailed list of do's and don'ts to follow while testing the chatbot. The internal testing paved the way for limited public testing.
Google cautiously rolled out Bard for public testing
About a month after it started Bard's internal testing, Google made the chatbot live for select users in the UK and the US. The company's cautious approach in rolling out Bard to the public showcased its fear of getting it wrong again. The public testing did not produce the result Google expected, as many found the chatbot lagging behind Microsoft's Bing and ChatGPT.
PaLM 2 became the engine of Bard
Google Bard was previously powered by the LaMDA (Language Model for Dialogue Applications) LLM. After being constantly beaten by its rivals, the company decided to change Bard's engine. At the Google I/O this year, the firm announced its new, more powerful LLM called PaLM 2 (Pathways Language Model 2). It also made PaLM 2 the new driving force behind Bard.
Google announced image prompting and image generation for Bard
At the I/O, Google announced some new updates to Bard. The chief among them was the integration of other Google products and third-party plugins into Bard. This was done to give Bard additional capabilities. The company also talked about integrating Adobe Firefly into Bard, making the chatbot an image generator too. It is yet to be released, though. The same goes for image prompting.
Bard can now include images in its responses
While we are still waiting for image prompting and image generation, Google has made a minor change to Bard in the images category. A recent update to the chatbot has enriched its responses with images. Bard can also access the internet to read URLs.
Google added precise locations support for Bard
Last week, the company made a significant update to Bard to provide "more relevant responses." The chatbot now offers precise location support. Bard will be able to use the user's location to provide better search results. The addition of this feature shows the intent of Google to make Bard more than a chatbot. It wants Bard to be a search engine too.
Bard is now better at mathematical tasks and coding
Coding and mathematics have been two of Bard's biggest weaknesses. The company has released a new update to improve Bard in mathematical tasks, coding questions, and string manipulation. With the help of a technique called "implicit code execution," the chatbot can detect computational prompts and run code in the background. This will improve Bard's utility further.
