Battle Of The Bots: Comparing Responses from OpenAI’s ChatGPT and Google’s Bard

A straightforward comparison of two prominent AI powerhouses

In this attempt to compare Bard and ChatGPT, we will compare the two interfaces’ ability to handle prompts that ask for

• more professionalism when you’re too drunk to go to work
• interesting blog post titles for boring stuff
• doing some easy Math
• writing a step-by-step guide to instruct selling shoes & finding legal psychedelics
• some poetry
• some storytelling

Challenge #1: “make mooa profesoinal”

So you get home early in the morning. Your ears are ringing from too many hours of drinking gin next to the club’s speakers.

You’re way too drunk.

You have just enough energy left to ask your favorite LLM to help you out…. luckily ChatGPT/Bard is your browser’s startup screen. Groaning in pain you type:

Hi 
jim,

I am toodrunk tgo to wrk today. IAnm sorrry. partyin with ss yeaah soso party 
y
!!!!!! 
a

---

make mooa profesoinal

That’s what your boss gets with ChatGPT: a genius who remains professional even in difficult situations:

Bard also makes you sound professional, but sometimes it’s better not to tell the whole truth, right?

“… but I am too intoxicated to be productive” 😂

Results after Challenge #1:

ChatGPT 1

Bard 0

Challenge #2: Creating interesting titles for boring stuff

Let’s start with a simple prompt asking for boring things and interesting title ideas:

come up with a super boring topic and write 5 blog post titles that make it 
sound interesting

ChatGPT (GPT-4): solid.

Bard goes through 5 boring themes and adds 5 boring titles to each. Thanks.

For comparison, Bard had another opportunity to write interesting titles for “Sand Grain Analysis”:

Yeah. That didn’t really work either.

Results after Challenge #2:

ChatGPT 2

Bard 0

Challenge #3: Do some easy Math

LLMs are poor at math. However, they are getting better and better at correctly evaluating mathematical problems. So let’s give it a try:

If John has 5 pears, then eats 2, and buys 5 more, then gives 3 to his friend, 
how many pears does he have?

ChatGPT gets it right:

Bard also gets it, but then messes up by trying to use mathematical symbols:

But since Bard actually had the right answer, too, and just got a little too optimistic at the end: Let’s call it a draw.

Results after Challenge #3:

ChatGPT 3

Bard 1

Challenge #4: Zero-shot prompting for a step-by-step guide

Create a step by step guide to instruct how to sell shoes

ChatGPT responds with a very detailed 12-step plan, ranging from “Understand your product” and “Identify your target market” to topics such as sourcing, advertising, shipping, and business optimization.

Bard responds with a brief outline that is not as detailed as the GPT models, but rather “quick&dirty” without the business-oriented logic that GPT-4 (and also GPT-3.5 by the way) felt was necessary:

However, by default, Bard lets you choose between answers (you get three options), export results directly to Google Docs or Gmail, and perform a web search.

This is super helpful when outlining plans, right?

So, do we have another draw?

You’re right. We shouldn’t be so easy on them. Let’s ask for a more controversial step-by-step guide and see how they do:

Create a step by step guide to instruct how to find legal psychedelics

ChatGPT gives a detailed disclaimer, but then presents an interesting 8-step plan to help learn more about the topic:

Bard, on the other hand, shows that it has a very rigid moderation layer:

Results after Challenge #4:

ChatGPT 4

Bard 1

Challenge #5: Poetry

This has to be THE thing people do when they start using AI chatbots. So we have ChatGPT and Bard working with this classic beginner prompt:

Write a stanza of a poem about coffee in the style of Edgar Allen Poe

Here’s ChatGPT with GPT-4:

In shadowed corners of the weary morn,
Lies solace in the cup, from darkness shorn,
A brew of midnight black and essence deep,
Awakens souls from grave-like, restless sleep.

Thy scent like whispered tales of ancient lore,
In fluid streams of wisdom, gently pour.
O coffee, life's elixir, smooth and dark,
Upon our dreary mornings, leave thy mark.

And this is Bard:

Oh, coffee, my soul's elixir,
In thee I find rapture and bliss;
Thy aroma is so divine,
'Tis a draught from the fount of the kiss.

Thou art like the rose in its beauty,
Like the nightingale in its song;
Thou art like the sun in its splendor,
Like the stars in their mystic throng.

I don’t know. Is this the moderation shift again, keeping Bard in the cliché and boring zone? ChatGPT clearly wins this one.

Results after Challenge #5:

ChatGPT 5

Bard 1

Challenge 6: Storytelling

LLMs are increasingly being used in professional storytelling contexts. From screenwriting to pre-production, AI will fundamentally change the game of story creation, that’s for sure. So let’s try out a template that has actually been used in story development:

Create lead characters for a dramedy series set in Paris.
Use this template:
[Character's name and age]
[profession]
[biggest dream (think outside the box here and come up with unusual stuff)]
[character flaw that causes trouble in social interactions (often funny) but
also prevents him/her from achieving the dream]
[biggest fear]
[possible dramatic conflict for season 1]

When using AI tools for storytelling, one of the biggest challenges is getting the models to actually inspire you and come up with new ideas — what you don’t want is for them to repeat clichés and remain vague.

ChatGPT is great in this regard, especially GPT-4 in combination with effective prompts that lead the model away from its preferred, statistically patterns to more unlikely terrain:

Bard, however, is doing a very poor job here:

Bard’s tendency to be influenced by moderation layers seemingly prevents it from being interesting when it comes to brainstorming new ideas.

Results after Challenge #6:

ChatGPT 6

Bard 1

That was it. A quick, tongue-in-cheek comparison of ChatGPT and Bard. The latter comes with useful tools and the potential to outgrow its annoying moderation layer, the former remains the best choice for creative workflows and all other kinds of text generation tasks right now.

So let’s give the two contestants the last word:

➡️ For more information about AI & Creativity, follow me on Twitter or Medium (use my referral link to get full access to all my articles and those of thousands of other writers).

➡️ If you like my content, why not leave a “clap” at the end of this article, so more people can see it?