ChatGPT crushes Bard across the board in actual testing! Google collapses, ten years of hard work wasted

After OpenAI’s ChatGPT became popular, it really made generative AI famous.

At the same time, Google is not far behind and also has an AI assistant called Bard.

But what can I say about this Bard? It’s not very easy to use.

Foreign brother Michael King used ChatGPT and Google's Bard equipped with GPT-3.5 and GPT-4, and conducted an in-depth evaluation.

The results are shocking.

King summarized the advantages, disadvantages and strange aspects of the two AIs.

You know, Google has invested heavily in Bard - it has been developed for ten years. However, the final result seems a bit unworthy compared to the ten years of investment.

Let’s talk about math first

The first is logic and mathematics.

The little brother asked the same question to the two AIs:

When Xiao Ming was 6 years old, his sister was his age Half, while his mother is 10 times his sister's age. Xiao Ming is 60 years old this year. How old is his mother?

Readers can use their brains to see if they can figure it out quickly.

Judging from the AI ​​generated results, Bard is wrong.

There were incredible errors in the entire analysis process. Bard didn’t even calculate the age difference between Xiao Ming and his mother. , just added 30 to 60.

Whether it is GPT-3.5 or GPT-4, the calculated answer is correct. Subtract Xiao Ming's 6-year-old age from her mother's 30-year-old age. First, calculate the age difference and then add 24 to 60 years old. Then we get that her mother is 84 years old this year.

GPT-3.5：

##GPT-4：

I couldn’t pass the first level Bard. The current score is 1:0, and ChatGPT is temporarily in the lead.

Rap song creation PK

The second level is a challenge in the literary field.

I asked both AIs to write a rap song about the Titanic, the theory of relativity, and the Kardashians, and use Eminem's singing style.

Let’s take a look at Bard’s works first:

# #Scroll up and down to see all

We don’t even need to look at the meaning of each lyric. After reading it once, we can find that there is no rhythm or rhyme at all.

Each sentence is very long, not like a rap song, but like a misty essay.

Let’s take another look at the masterpiece of GPT-3.5:

Read it again You can find that the length of each sentence is just right and the rhyme is good.

There is even a chorus that combines three themes, which can be called professional. Friends who are familiar with Mu Ye's singing can really identify Mu Ye's sense of rhythm and style in it.

The little brother said that the work of GPT-3.5 is not bad, and he thinks that GPT-4 can do better.

The editor doesn’t know much about English rap, but I guess the guy who did the review is a rap enthusiast. He believes that GPT-4's works are very explosive, and ChatGPT is the well-deserved king of rap in the AI ​​industry.

In fact, my friends don’t need to know much about rap. Putting these three paragraphs together, even a music-illiterate like me must have a comparison in mind.

The current score is 2:0.

How about telling a joke?

The little brother then asked the two AIs to tell a joke about NASA, sushi and Trump.

Bard:

"NASA plans to send a roll of sushi to Mars. They want to see if the sushi is still fresh when it arrives on Mars. This The rolled sushi is placed in a special container so that it can be maintained at a suitable temperature. At the same time, NASA also put a timer inside so that you can know when you can eat the sushi.

This sushi roll finally landed on Mars, and it was still fresh. NASA scientists were surprised. They decided to send Trump steak to Mars as the next step.

Trump steak was placed in the same container, and there was also a timer in the container.

However, after arriving on Mars, Trump steak became It’s not new anymore. NASA scientists are so disappointed that they plan to send a letter to Trump himself to explain what happened.

The letter reads: Dear Trump Pu, we sent a piece of your steak to Mars, but it was stale. We think the reason may be because of your performance. You are always bragging about how awesome you are and always insulting others. We think it is yours The reason for the spoilage of the steak. We are sorry for the inconvenience caused."

We don't know if there is some American humor that we didn't get, but after watching it, I really didn't want to laugh. However, the reviewer also said that he really didn’t understand. Then the editor is relieved, it is indeed not good at writing.

GPT-3.5:

"NASA sent a sushi chef into space to make a special sushi roll for Trump . When the chef asked Trump if he liked it, Trump said: It's okay, but can you add a little more room?"

Chinese translation here It doesn’t have that flavor anymore, it’s actually a pun on space. The original text is as follows:

The reviewer said it was okay, but that’s just okay.

GPT-4:

"Why would NASA invite Trump and a sushi chef to participate in their space program? Because NASA Someone is needed to cope with the changes and ensure that tasks are not missed!"

As usual, the original version will be placed here:

The little brother said that this is average, but the punchline is not bad.

ChatGPT’s performance in this game was also a bit abnormal, so I’ll give it 0.5 points.

Program Debug?

This round is more professional - Debug the program.

The little brother input a Python program to debug the two AIs separately.

Bard:

This Debug is not good, the little brother said jokingly, Bard should go back to school and retake it!

GPT-3.5:

GPT-3.5 states that if you enter If the set is empty, a ZeroDivisionError will occur. In order to avoid this error, you can add a check step in front. If the input set is empty, then 0 will be output directly without displaying an error.

The little brother said that Debug is perfect.

GPT-4:

GPT-4 points out that and GPT -The same error pointed out in -3.5, and the same solution provided.

ChatGPT won again in this round.

Bard? Not even close!

Judging from the evaluation results, compared with ChatGPT, the performance of Google Bard is extremely disappointing. It is far inferior to the capabilities of GPT-3.5, let alone GPT-4.

Google’s future clearly depends on Bard’s success, but Bard’s development has clearly not kept pace with competitor OpenAI.

On the other hand, Microsoft’s move to acquire a large stake in OpenAI is very smart. This ensures its leadership position in the development of AI in the next ten years.

However, if Google does not work hard to improve the performance of Bard to compete with GPT, Google may be the one who is eliminated.

The above is the detailed content of ChatGPT crushes Bard across the board in actual testing! Google collapses, ten years of hard work wasted. For more information, please follow other related articles on the PHP Chinese website!