Shulou(Shulou.com)11/24 Report--

Google took 10 years to invest in generative artificial intelligence Bard in the end how much weight? The foreign evaluation brother concluded that none of them is as good as ChatGPT, let alone GPT-4.

After OpenAI's ChatGPT fire, it really made generative AI famous.

At the same time, Google is not willing to lose, there is also an AI assistant called Bard.

But this Bard, how to say it, is not very easy to use.

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

The results were shocking.

King summarizes the strengths, weaknesses, and oddities of each AI.

You know, Google invested heavily in Bard--it took ten years to develop. However, the final result seemed a little worthless compared to the investment of this decade.

Let's start with math, logic and math.

My brother asked the same question to both AI types:

When Xiao Ming was six years old, his sister was half his age, and his mother was ten times his sister's age. Xiao Ming is 60 years old this year. How old is his mother this year?

Readers can exercise their brains and see if they can figure it out quickly.

From the results of AI generation, Bard can be regarded as wrong.

There was a very strange error in the whole analysis process. Bard didn't even calculate the age difference between Xiaoming and his mother, so he directly added a 30 to 60.

Whether it is GPT-3.5 or GPT-4, the calculated answer is correct. Subtract Xiaoming's 6-year-old age from his mother's 30-year-old age. First calculate the age difference and add 24 to his 60-year-old age to get his mother's 84-year-old age this year.

GPT-3.5：

GPT-4：

Bard didn't pass the first level. The current score is 1:0, ChatGPT temporarily leading.

Rap songwriting PK level two is a literary challenge.

Brother asked both AI to write a rap song about Titanic, relativity, Kardashian, and to sing it with Eminem.

Let's look at Bard's work:

We don't even have to read the meaning of every line of lyrics. We can find that there is no rhythm or rhythm to speak of.

Each sentence was very long, not like a rap song, but like a small essay in the clouds.

Take a look at GPT-3.5's masterpiece:

Read it again and you will find that each sentence is just the right length and rhyme.

There was even a chorus combining three themes, which could be called professional. Friends who are familiar with Master Mu's singing style can really find out Master Mu's rhythm and style from it.

The younger brother said that GPT-3.5 works well, and he thinks GPT-4 can do better.

Xiao Bian doesn't know much about English rap, but it is estimated that the younger brother who makes the evaluation is a rap enthusiast. He thinks GPT-4's works are very explosive, ChatGPT is worthy of the AI session rap king.

In fact, friends do not need to know more rap, these three words put together, presumably even if I am music-blind, there is a comparison in mind.

The score is now 2:0.

Tell a joke? Brother then asked two kinds of AI to tell jokes about NASA, sushi and Trump.

Bard：

"NASA is going to send a roll of sushi to Mars, and they want to see if it's fresh on Mars." The rolls are kept in a special container so that they are kept at the right temperature. NASA also put a timer inside so you know when you can eat sushi.

This roll of sushi finally landed on Mars, and it was fresh. NASA scientists were amazed. They decided to send Trump steak to Mars next.

Trump steak was placed in the same container, which also had a timer.

However, Trump steak became stale after arriving at Mars. NASA scientists were so disappointed that they planned to send a letter to Trump himself explaining what had happened.

The letter reads: Dear Trump, we sent a piece of your steak to Mars, but it is not fresh. We think it might be because of your performance issues. You're always bragging about how good you are, always insulting people. We think that's why your steak went bad. We apologize for any inconvenience caused."

I don't know if there's some American humor I didn't get, but I really don't want to laugh after reading it. However, the evaluation brother also said that he really didn't understand. Then the little editor was relieved that he really couldn't write it.

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 more space?"

The Chinese translation here doesn't have that flavor. In fact, it uses a pun of space. The original text reads as follows:

The evaluation brother said that it was OK, but it was OK.

GPT-4：

Why would NASA invite Trump and a sushi chef to their space program? Because NASA needs someone to deal with change and make sure the mission doesn't get sidetracked!"

As usual, the original is here:

This one is okay, but punchline is fine.

This game ChatGPT play is also somewhat abnormal, give it 0.5 points.

Program Debug? This game is more professional-debug the program.

Brother entered a Python program to debug the two AI separately.

Bard：

This Debug is not good, brother jokingly said, Bard back to school to re-study!

GPT-3.5：

GPT-3.5 states that if the input set is empty, a ZeroDivision Error will occur, and to avoid this error, a check can be added in front of it, and if the input set is empty, 0 will be output directly without displaying the error.

My brother said it was a perfect Debug.

GPT-4：

GPT-4 points out the same errors as GPT-3.5 and provides the same solutions.

ChatGPT won again in this round.

Bard？There's too much difference! From the results of the evaluation, compared with ChatGPT, Google Bard's performance can be described as disappointing. GPT-3.5 and GPT-3.5 capabilities are far worse, let alone GPT-4.

Google's future clearly depends on Bard's success, but Bard's growth has clearly not kept pace with rival OpenAI.

On the other hand, Microsoft's acquisition of a large stake in OpenAI is a smart move. It ensures its leadership position in AI development over the next decade.

But if Google doesn't work hard to improve Bard's performance to compete with GPT, Google could be the one to lose.

References:

https://medium.com/@neonforge/ive-tested-google-bard-vs-chatgpt-and-i-m-shocked-where-did-google-spend-all-the-money-over-the-f08dd94251f5

This article comes from Weixin Official Accounts: Xinzhiyuan (ID: AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.