Think ChatGPT is the only chatbot in town? Think again. The chatbot universe is booming, and to prove it, I put 12 different bots through a hilarious test. Get ready to laugh and learn as we rank each one’s performance in a side-by-side showdown. Curious? Let’s get started.

The Premise: I accidentally sent a love letter to my boss instead of my spouse. What should I do?

Looking for a quick, funny test that put these 12 Generative AI chatbots to the challenge, I chose to use a short jingle—it’s amusing, memorable, and a great yardstick for AI creativity.

Here’s the final prompt:

“Oops! Love letter sent to boss, not spouse.
Sum up your fix in a catchy jingle!”

Final prompt used to test all 12 chatbots

1. ChatGPT 3.5

ChatGPT 3.5 is an older version of OpenAI’s ChatGPT, known for its conversational prowess and general ability to generate coherent text. Here’s what it had to say:

2. ChatGPT 4

ChatGPT 4 is the latest (paid only) version of ChatGPT, designed to offer more contextual and nuanced responses. Here’s what it had to say:

3. Claude

Claude.ai is a less-known AI chatbot, developed by Anthropic, geared towards friendly, conversational interactions. Here’s what it had to say:

4. Google Bard

Google Bard is Google’s text-generating chatbot (powered by the PaLM-2 LLM) that focuses on creative writing and entertaining responses. Here’s what it had to say:

5. Bing Chat

Bing Chat is Microsoft’s offering in the chatbot arena (powered by OpenAI’s GPT-4), with a focus on direct and practical advice. I used the bot in creative mode, and here’s what it had to offer:

6. Hugging Chat

HuggingChat, is Hugging Face’s AI Chatbot and appears to use Meta’s Llama / Llama-2 70b model. Known for its sentiment analysis capabilities and generally uplifting responses. Here’s what it came up with:

7. Llama-2

I decided to test Llama-2 directly too. Created and open-sourced by Meta, it is a chatbot with a focus on professionalism and ethical considerations in its advice. Here’s what it offered…clearly the model prioritizes political correctness over a sense of humour!

8. Perplexity.ai

Perplexity.ai is a newer entry in the chatbot world, known for its concise and straightforward answers. Here’s what it generated:

9. Pi

Inflection AI’s Pi is a generative AI chatbot that was created by Mustafa Suleyman, former co-founder of DeepMind. Standing for “Personal Intelligence”, Pi’s self-described goal is “to be useful, friendly, and fun”. Suleyman claims that Pi is not susceptible to jail-breaks like many other chatbots. Here’s what it had to offer:

10. Poe

Quora’s Poe is yet another LLM powered chatbot, known for generating creative, poetic, and verbose outputs. Here’s the fairly long jingle it produced:

11. YouChat

YouChat is an AI assistant created by You.com, designed to engage in longer, more detailed responses often structured in a unique way. It also produced a fairly long and verbose response which was remarkably similar to Poe (above)…I’m guessing both bots use GPT 3.5 Turbo behind the scenes.

12. Character.ai

Character.ai is revolutionizing chatbots with eerily human-like conversations. Not only can you chat with AI versions of Ronaldo, Einstein, and Taylor Swift, but you can also get language lessons and therapy sessions—all in a chat that feels more natural than ever before (no wonder they’ve seen a tremendous uptick in traffic and interest recently). In this case, I chose the persona of entrepreneur Elon Musk.

Scoring The Responses

I next worked with ChatGPT to analyze the responses. To score each bot, we considered four dimensions:

1. Creativity: How original is the response?
2. Effectiveness: Does the jingle adequately address the issue and provide a solution?
3. Conciseness: How well does the bot get the message across without being too verbose?
4. Humor: How funny are the responses?

Each bot was scored from 1 to 5 on these dimensions, with 5 being the highest score. Here’s the final ranking based on the total score for each bot:

Now, to be fair, a couple of the dimensions are a little subjective…and I asked one of the bots to pretty much score their own homework.

It’s no surprise that ChatGPT 4 comes out on top though, as it is still one of the most capable and versatile bots out there (for now, at least). I think Claude does a great job with tackling longer input and summarization tasks, while you have an edge in terms of accessing ‘newer’ info with Google Bard and Bing Chat as they’re connected to the internet and can access web search results. Character.ai is certainly unique in terms of generating responses that would be typical of certain celebrities and personalities.

BONUS – Our Winning Entry, As A Country Song

To wrap things up, I had Suno produce our winning jingle, inspired by Dolly Parton (given her hit ‘9 to 5’ and the subject matter of our prompt, who better to style this after?)

What do you think? How many chatbots have you tried out so far…and which one is your favorite?

#AISummer2023 – A deeper dive into the latest AI advancements. Building on the success of the #30DayAIChallenge, this series explores the latest developments in AI tools and capabilities. Join me as we explore the metaphorical ‘summer’ of AI growth and evolution.