Bard: Not ready for prime time, yet!

Why do I think that Google should have waited longer before releasing it?

Google released Bard on 21st Mar 2023 (2 days ago). I immediately joined the waiting list. I got the approval just an hour ago. I tried it for a few different tasks and the experience has been worse than underwhelming.

CODING

Verdict: Fail

I am a programmer, and my first question was about C# project I have been working on. It generated basic sample code. It was subpar but it could compile. As I dug deeper into that code with Bard, it responded.

It looks like you want my help with coding, but I'm not trained to do that,
yet. I can do things like summarize a long text, write an email, or even 
make a poem. Is there something else you would like to try?        

It is a failure from my perspective. It might be able to do some kind of coding work but not as a reliable partner, especially not with someone who has experience.

POEM Writing

Verdict: Average

I enjoy reading/writing poetry and thus I wanted to know how good of a poem Bard could write. I will not share the generated poem here due to the sensitive topic I had chosen. However, I would say that the content was okay but rhyming was missing. I am not sure if its default style is rap but I did not appreciate the lack of rhyming.?The choice of words and sentences also seemed mechanical and lacked the human feel.

PROSE

Verdict: Fail

I asked Bard to draft an essay on “Cow” and to keep it in Indian context.

Bard generated an essay that was mediocre quality. It did not cover the aspects comprehensively and sentences were mechanical and repetitive.

Look at the sample below. You see how the same sentence is repeated multiple times (as if it ran out of ideas).

TRAVEL PLAN

Verdict: Fail

I asked Bard to plan a trip for a family, traveling from Metuchen, NJ to Orlando, FL. I instructed it that family intends to drive their car.

Here is the first line:

Day 1
Leave Metuchen in the morning and drive to Orlando. The drive should take 
about 10 hours.        

It says 10 hours. The distance between Metuchen and Orlando is 1051 miles. I am not sure how Bard expected me to drive but even Google Maps says it will be around 16 hours. I wonder if Bard is not using Google Maps to make these decisions.

FACTUAL

Verdict: Fail

I was under the impression that bard might be good in coding. I decided to try my luck again hoping that the previous issue might have been just a one off. This time, I made a simple code request and asked Bard to write a sample code to use Bard in programming.

Bard did not bother to ask which language I would like the code in. Well, the above code is python. I asked Bard if it was sure that "bard" package is available in python that can be used.

Bard responded:

No, there is no Python library called 'bard'. I apologize for the 
confusion I am still under development, and I am always learning new 
things. I will try my best to be more accurate in the future.        

At this point, I had almost given up on Bard. However, as a parting shot, I asked it to compare Bard with chatGPT. It wrote this.

...
..Bard is still under development, and it is not as mature as ChatGPT...
...        

Well, even with all the flaws, I should say that Bard is candid about its own inferiority. That is a plus point.

Google CEO Sundar Pichai had told employees that 80,000 of them 
helped test Bard A.I., while warning "things will go wrong"...

I wonder if it is the "confirmation bias" that might result in 80,000
testers finding the quality (as shown above) acceptable. I spent less 
than an hour and found enough issues to find it unfit for public use.