[Review] Is Google's Bard AI chatbot better than Bing & ChatGPT?
In this review, we compare Google's AI chatbot, Bard, to Bing and ChatGPT in a few different use cases. Bard is described as a "creative and helpful collaborator" that allows users to collaborate with generative AI, and is based on the same technology behind LaMDA. One unique feature of Bard is that it offers three different drafted answers per prompt and allows users to provide feedback on each of them.
When asked about the character count of the sentence "Once upon a time, there was a girl with red shoes," Bing and ChatGPT both struggled to provide the correct answer. Although Bing's answer was close, it failed to count spaces. ChatGPT initially forgot to include spaces in its answer, but was able to derive the correct number after being prompted. However, Bard provided various answers that were all incorrect. When told the real answer, Bard provided a confusing explanation that left the user perplexed.
For the prompt to write a story about a black cat named Cocoa who lives in an office, Bard, Bing, and ChatGPT all came up with unique stories. Bing went for a fantastical route while ChatGPT's story was cute but slightly horrific. Bard's story was incredibly adorable and stood out by being the only one to correctly assume that Cocoa is a boy. However, the writing style was more storybook-esque compared to ChatGPT's writing, which sounded more human and mature.
Finally, to test the AI chatbots' ability to access the internet and stay up to date with news, the prompt asked about today's startup news in Malaysia. Bing was able to provide a curated answer with relevant sources, while ChatGPT suggested several publications for the user to explore. However, Bard had to admit that it did not have access to real-time news updates and only offered resources for finding startup news in Malaysia.
In conclusion, while all three AI chatbots had their strengths and weaknesses, Bard's performance was less consistent compared to Bing and ChatGPT. Its unique feature of offering three different drafted answers per prompt may be useful for users who want more options, but its responses were often confusing and incorrect.
