How does Google’s Bard compare to ChatGPT?

Over the weekend, I had the opportunity to explore Bard, Google's latest LLM contender against ChatGPT. As an avid user of ChatGPT-4, I could not resist spending too much of my weekend comparing the two. Here are my initial impressions.

?

What I like about Bard:

• Current information: The answers provided by Bard are up-to-date, and Google did a good job implementing online access and the comprehensive power of their search engine.
• Speed: Bard feels notably more responsive. It generates larger answers more quickly, eliminating the need to switch tabs and return later, as I often do with ChatGPT.
• Conciseness: Bard's responses are concise and focused, with less fluff than those of ChatGPT.
• Sources: Bard can provide sources for its responses, which simplifies fact-checking compared to ChatGPT (a crucial step when working with LLMs).

?

Where ChatGPT does a better job:

• Persistence: ChatGPT goes above and beyond to address your questions, offering alternatives when it cannot provide a solution (and sometimes even fabricating answers). Bard, on the other hand, is more likely to decline answering. While this boosts trustworthiness, it may hinder brainstorming.
• Iteration: ChatGPT demonstrates a stronger understanding of context and excels at refining drafts. Bard tends to incorporate new context into the existing message rather than generating an entirely new draft, leading to an unpleasantly disjointed, bullet-point style text.
• Placeholders: ChatGPT more effectively utilises placeholders, generating drafts and prompting you to fill in the blanks when context is lacking. I was not quite as satisfied with Bard’s use of placeholders, which often generates a more generic message instead.
• Tone: ChatGPT's responses are more sophisticated in their wording. Comparing letters drafted by ChatGPT and Bard, one might assume ChatGPT possesses a higher level of education.
• Coding: ChatGPT significantly outperforms Bard in generating and explaining code. Bard occasionally even refuses to generate code. And while OpenAI’s API is powerful and easy to implement, Bard itself could not explain to me how to use its API.?

?

Verdict:

Personally, I use ChatGPT a lot for its text drafting and coding capabilities. So far, I like Bard's answers for everyday searches or messages since they are more to the point and strike a better balance between fluff and content.

However, I find ChatGPT more effective at handling complex responses that demand extensive argumentation, particularly after the recent upgrade to GPT-4 (e.g. customer service inquiries, such as airline ticket fee refunds under EU regulations). Of course, both ChatGPT and Bard supported me in drafting this article, but I mostly favoured ChatGPT’s recommendations. Especially ChatGPT’s coding capabilities are unmatched by Bard, which often outright refuses to help you.

In conclusion, both are incredibly powerful, each possessing distinct advantages in specific areas. Based on my personal and anecdotal experiences, Bard feels like a next-generation search engine, while ChatGPT resembles a personal assistant. However, what excites me the most about Bard is that it presents a formidable competitor for ChatGPT, promoting competition and continuous innovation in the industry. The best is yet to come – what a time to be alive!

?

Did you have similar or contrasting experiences? Share your thoughts in the comments!