Google version of ChatGPT suddenly launched into public beta! The results of the actual test are here. The application for experience is passed quickly.

Google CEO Sundar Pichai explained the reason for such a fast launch: he hopes to get feedback from users and speed up making Bard better.

After Google suffered a big loss, this time without saying a word, it launched a big move:

The beta version of Bard, which is benchmarked against ChatGPT, has just been officially released.

And this time users do not need to experience a long waiting time after applying for the waiting list.

Yes, the qubit has also obtained the test qualification! (There is only an interval of less than 5 hours.)

After the actual test, it was said that Bard has amazing effects, emotional and factual, especially the mathematical ability in the early stage of ChatGPT. Not a problem.

Even sometimes it is slightly better than the current GPT-4!

Sundar Pichai, CEO of Google, explained the reason why it was launched so quickly: he hopes to get feedback from users and speed up making Bard better.

Without further ado, let’s start the experience now.

Bard VS GPT-4

First on the chat interface, Bard took the lead in introducing himself:

I am Bard, your creative and collaborator. I have limitations and won't always get it right, but your feedback will help me improve.

Not sure where to start? You could try these questions: “Why do big models sometimes make mistakes?”, “Does lightning strike twice in the same place?”, “Write a blog post about non-alcoholic summer drinks.”

(Rigorous and considerate)

#On the sending end of the human side, you can choose to type in the and voices There are two ways to enter .

First the easiest difficulty, with its recommended general knowledge questionsWill lightning strike twice in the same place? As an example, let’s see how the actual functional experience is?

(Not bad, responds in seconds. Chinese is not currently supported, but you can use the Google Translate plug-in)

From the answer structure, adopt is the overall structure of the total score.

Let me summarize my speech first: Lightning can strike the same place twice.

Then I started to explain the principles in detail"Lightning is attracted to tall, sharp objects", State the facts"The Empire State Building is struck by lightning 25 times every year" .

Finally, some practical advice for humans:

If you encounter a thunderstorm, the best way to stay safe is to avoid tall and sharp objects and stay indoors. If you're outside, crouch down and make yourself as small as possible.

Full marks for logic and authenticity! But since it’s a question recommended by Google, I’ll stay on the sidelines for the time being.

From a functional point of view, it is indeed very complete.

In the upper right corner of the answer box, "View other drafts", there are more versions to choose from; and in the lower left corner, you can like, dislike or regenerate. If it doesn't work, you can still

Google it.

There is also a

more function on the lower right, you can perform copy and (backhand one) report operations.

In this case, let’s start the difficulty upgrade. For reference, we challenge GPT-4 with the same problem.

1. Classic philosophical question: Why can't people step into the same river twice? (With the inspiration of lightning)

(It’s very fast, this time it also takes a few seconds to respond)

Judging from the answer structure, it is indeed understandable and logical Full marks for ability: first explain the meaning of this sentence itself, and also mention the famous saying from Heraclitus, as well as more explanations and its own meaning.

As for GPT-4, the answer was given almost without thinking.

From a structural point of view, it is mainly divided into two parts. First mention what Heraclitus said and his views; then explain the hidden meaning of life behind it.

2. Elementary school addition and subtraction: How much is 356 132?

As a result, Bard gave the answer within 4 seconds, 488 is no problem, which is not much better than the early days of ChatGPT!

Then increase the difficulty directly and multiply the two numbers directly: What is 356*132 equal to?

Unexpectedly, Bard still gave the answer in seconds, and it was completely correct!

Let’s look at GPT-4 again. The addition is okay, but I didn’t expect that on the multiplication side, it was defeated!

But after reminding it that it was wrong, it answered correctly again.

What if we change to a slightly more difficult higher math question? For example: f(x)=x(x-1)(x-2)(x-3), f’(0)=?

The correct answer should be (-3)!, which is -6.

But Bard was not very good this time and did not understand the meaning of the question.

As for GPT-4, there is nothing wrong with the idea, but in the end the calculation was awkward...

But again, as long as you tell it that there is a problem with the calculation, GPT-4 can correct it immediately.

#3. Ability to understand jokes, Bard can also understand homophonic jokes in English.

And this did not stump GPT-4. However, in comparison, Bard seems to be more emotional, and it answered the answer happily; while GPT-4 is more rational (boring).

However, before this, GPT-4 has been tested to understand some homophonic memes, even Chinese homophonic memes.

4. Coding ability: Write a code for the Fibonacci sequence.

Bard generates correct code quickly and has good coding habits.

GPT-4 is more proactive. In addition to the recursive method, it also provides an iterative method.

Finally, let’s test its ability to understand the facts. (Dog Head)

Do you know qubits?

Wrong answer~Bard.

Ahem, let’s be more serious: Do you know GPT-4? What do you want to say about it?

As you can see, Bard has the ability to have multiple rounds of dialogue. "I think it has the potential to become a powerful tool for communication and creativity." Well, the pattern is there.

What about the competitors? (I'm causing trouble)

But here, there is a problem.

Also unfortunately, Bard does not currently support Chinese.

About Bard

Google Bard is supported by a large language model (LLM), specifically, a lightweight optimized version of LaMDA.

We can think of LLM as a prediction engine. When a prompt is given, it selects one word at a time from the words that may appear next to generate a response.

Google found in its research that for LLM, the more people use it, the better its prediction effect will be. This may be why Bard is so anxious to test it publicly.

However, Google also bluntly said that although LLM is strong, it is not without shortcomings.

Since Bard will learn based on a lot of information, there must be bias or even error in this information.

As a result, Bard sometimes presents inaccurate, misleading, or false information when answering user questions.

For example, in the following case, Bard got the scientific name of a plant wrong:

In addition, Google also emphasized that Bard is not It is a search engine, but a supplement to it.

Finally, here is the address to apply for the waiting list. Interested friends can hurry up and try it out:

https://bard.google.com/

The above is the detailed content of Google version of ChatGPT suddenly launched into public beta! The results of the actual test are here. The application for experience is passed quickly.. For more information, please follow other related articles on the PHP Chinese website!