Google Bard. An AI tool that completes and competes with ChatGPT.

According to the dictionary, a Bard is "a tribal poet-singer skilled in composing and reciting verses on heroes and their deeds". But I assume you are here to read about Bard, Google's experimental conversational AI service powered by LaMDA. After testing ChatGPT around four months ago I had the chance to try Bard today. As my first approach, I used 4 different questions and compared the results with ChatGPT 3.5. The outcome was interesting, as expected, and it seems that the AI revolution has a long road ahead.

We mainly use AI to help us with homework or work. So my first question came from classics. I am not an expert in Latin, so there is a 50/50 chance that AI would give me a bad mark.

I will need some help from people in Classic Twitter to get the winner here since the two answers are entirely different. I like the fact that Bard provides different draft versions, although it seems they are very similar to each other. You can also have the option to "Google it", which does not make much sense for a Google product that should have used the power of the search engine.

Then I tried Bard on coding. Since I wanted to go a bit further I tried it with two different languages, Python and COBOL.

For the Python test, I chose ChatGPT as the winner, even though each tool has a different approach with different benefits. The code produced by ChatGPT looks better. I like the comments and the style of the output, plus it's written as a code snippet that I can paste to any other programme. On the other hand, Bard gave a proper Python programme with a main function and better structure. I think more experienced programmers would like the proper style of Bard and that leads us to the next coding test: legacy code.

Which AI tool can help with a COBOL task? Probably Bard. 

Bard is a clear winner on this test. Although COBOL is not the most commonly used programming language I would like to have a tool that can help me with as many problems as possible. ChatGPT provided the algorithm for the Fast Fourier Transform (FFT) but the task of translating that to COBOL would be equally difficult.

We know that ChatGPT does not have access to the newest information, so maybe that's the winning point for Bard. When I asked for a generic question like the population of Greece, Bard provided a more recent answer and it listed the links to the sources, something that ChatGPT does not (yet).

 

As the final test, I wanted to know if each tool is aware of the competition and it seems that they are. 

It is interesting that ChatGPT is "aware" of Bard while Bard is "familiar" with ChatGPT. 

Let the games begin!