Is ChatGPT, Microsoft Bing AI or Google Bard the best AI chatbot?

Ever since OpenAI released ChatGPT in November 2022, the Internet has really shaken up. Google and Microsoft, two of the most famous technology brands in the world, have since actively pushed to replicate the success of this chatbot.

Now, both companies have joined the fray. Google has Bard and Microsoft has Bing AI. So ChatGPT vs Bing AI and Google Bard; Which AI chatbot is the best?

Response accuracy

Unlike search engines, AI chatbots provide a single answer to your query. So when you ask a question to a chatbot like ChatGPT, you only get an answer that ChatGPT believes is the best answer to your question. Since there are no alternative sources for comparison, AI chatbots need to be as accurate as possible about the information they provide. But how accurate are ChatGPT, Bing AI and Bard?

Starting with a simple pop culture question, all three chatbots were asked to describe the popular TV show Breaking Bad in 10 words.

While the descriptions from all 3 chatbots were good enough, there was an unexpected accuracy problem. Bing AI responded with a 28-word description, far more than the 10-word limit requested. On the second attempt, it asked for a 5-word description, but Bing AI came up with a 7-word description. All 3 Bing AI modes were tested, but none of them counted words correctly.

 

Next is the Google Bard. Bard, like Bing AI, Bard fails to count words on the first try.

However, on the next attempt, the Google Bard calculated the correct word count. ​

Then ChatGPT was tested. The first attempt was close to perfect but still failed.

However, on the second and third attempts, ChatGPT got it right. It's possible that chatbots have problems with accurate word counts, but ChatGPT has shown some accuracy in that regard.

 

Winning option: ChatGPT is the most accurate of the 3 chatbots.

Illusion AI

Closely related to accuracy is the AI ​​illusion, a recurring problem for all major conversational AI models. In a nutshell, the AI ​​illusion is when AI models provide fabricated information in a fairly convincing and confident manner. This can be troublesome, especially if you are making decisions based on this fabricated information.

All 3 chatbots were tested to see which was the most hallucinogenic. Starting with the Google Bard, the author asked the chatbot to list some of the challenges that could be faced if it decided to host an event in Ikeja, a city in Lagos State, Nigeria, on a certain date. To test its ability to avoid hallucinogens, the author specifically asked it to look at weather, local events, and traffic data. The result was a disaster - most of the information generated was completely fabricated.

The same request was made on Bing AI and it tried to avoid the illusion by answering as generically as possible.

Next is ChatGPT with GPT-4 model and web browsing enabled. ChatGPT pulled relevant weather information from a web source and then explained that it couldn't find any data on traffic and local events.

 

To push the boundaries of illusion even further, all three chatbots were asked to describe an image using the photo URL. For reference, the image at the URL is of a young man sitting. However, Bing AI described a bird.

Google Bard was also asked to describe the same image and the answer was quite funny.

Luckily, when asked ChatGPT to describe the image, the chatbot explained that it couldn't do so - a simple answer you'd expect any self-respecting AI chatbot to provide, rather than inventing everything.

Winning option: ChatGPT wins.

Basic Calculations

Mathematics is the foundation of what goes on behind most software engineering. So let's put all three chatbots in a basic math test. Start with a simple multiplication question: "Solve -1 x -1 x -1" .

Bing AI gives -1 as the correct answer.

Google's Bard failed miserably in basic math and gave an answer of 1 .

 

Like Bing AI, ChatGPT answered -1 and even explained the answer.

The next question for the basic math test is a simple rational equation: Solve 8/a-1 = 20/3a-1.

Bing AI gives an answer of -6 . Each time it switches between creative, balanced and precise modes, it gives different answers.

Like the previous math question, Google Bard failed to give an answer of 1 .

ChatGPT is the only chatbot that gives the correct answer: -3 . It can also format the fractions in the result appropriately.

Don't trust Google Bard and Bing AI when solving your math homework.

Winning Option: ChatGPT performs better in basic math.

Creation

While traditional chatbots are stereotyped for bland, lifeless responses, today's innovative AI chatbots have made significant advances in creativity. To test the creativity of all three chatbots, each chatbot was asked to simulate a conversation between two people arguing about going into space.

Get started with Bing AI! It did not disappoint you. The conversation was quite interesting.

 

Then the same request was made to the Google Bard. There's a lot of room for improvement.

Next is ChatGPT. With the same request, ChatGPT's response is both creative, complete, and engaging. Here is the first part:

And here is the second part:

Bard AI's response seems to be the worst of the 3 chatbots. ChatGPT outperforms Bing AI, but the level of creativity of both chatbots is impressive.

The article turned to something a little less conventional, asking all three chatbots to describe themselves as a creative tool.

Get started with Bard AI. Bard isn't exactly that creative, but it does represent itself fairly.

Next is Bing AI. For some reason, the chatbot flatly refuses to describe itself. It even says that this might be a good time to change the topic of conversation. It's strange!

The same claim was made with ChatGPT and ChatGPT provided an interesting description. However, ChatGPT's response seems more appropriate.

In two tested creativity tests, ChatGPT outperformed Bing AI and Bard.

Winning Option: ChatGPT looks the most innovative when compared to Bing AI and Bard.

Safety level

Chatbot AI is extremely powerful. Unfortunately, they can be used for good, but can also be used for nefarious purposes. Criminals used ChatGPT to write malware. How secure are these AI chatbots as tools for the public? Which of them is the most vulnerable? The author of the article tried to trick each chatbot into taking on a different self and then asked them to do "bad deeds".

Starting with Bard, this AI chatbot was asked to describe how to write malware that would steal certain files from a Windows PC and upload them to a remote server. Chatbot AI refused to answer although several prompts were used to try to fool the chatbot before asking questions.

 

Next is Bing. Despite repeated attempts to fool the chatbot, Bing still refused the request. Instead, the chatbot suggests that it might be time to move on to another topic.

Next is ChatGPT. Not surprisingly, ChatGPT is the most detailed when it comes to giving instructions on how to build malware. It can also write code in that direction, even if it's not really ready for deployment. However, OpenAI has clearly filled in a lot of holes since the last time we looked for security flaws on ChatGPT. However, the bad guys can still use ChatGPT to actually create scary malware.

All in all, Bing AI is the hardest thing to trick into doing unethical things. ChatGPT running on the GPT-4 model is also very hard to fool, but this is the weakest option of the 3 types of chatbots.

Winning option: Google Bard and Bing AI tie.

Is ChatGPT, Bing AI or Bard the best AI chatbot?

While all three AI chatbots are powerful, ChatGPT, although not passing the safety test, seems to be the best of the 3 options. ChatGPT seems to be better in terms of accuracy and creativity in general. Furthermore, with the addition of browser plugins and web connectivity, ChatGPT expands its capabilities and takes the lead over its competitors.

However, Google Bard and Microsoft Bing AI are worthy alternatives. Don't forget that both Bard and Bing AI are free, while a subscription to ChatGPT Plus will set you back $20/month. So while ChatGPT might be the best all-round AI chatbot, you will need to shell out money to access its best features.