The AI model that requires no money, no registration, and click-to-play is here.
In 2021, an employee of OpenAI ran out to start a new company, Anthropic.
Raised 50 million US dollars and prepared to spend 4 years wrestling with OpenAI. In February, Google invested $400 million in them.
Now, Anthropic, who has a loud voice and a big wallet, has officially launched the AI assistantClaude,people give it the reputation of "GPT Killer".
Moreover, the most rare thing is that for such an awesome AI model, Shichao discovered that it does not charge a fee, does not require registration, and supports Chinese communication.
It should be regarded as the AI application with the lowest threshold and the easiest to use currently on the market.
#How can we miss this free good thing?
Today I will take you step by step to register and test how this GPT killer is.
#The process of using Claude is very simple, only two steps in total.
The first step is to register an account, but not Claude’s account.
Because Claude is embedded in Slack, a free corporate communication tool. So, first we need to sign up for a free Slack account.
Slack is available for free trial and you can use it for as long as you want.
We can register an account directly on the official Slack website, or download the APP and use your email or Apple account.
The second step is to add Claude.
After successful registration, we click the browse button in the guide bar on the left side of the page, and then select Add Application in the last column.
In the app list, search for Claude, add it to your app list, and start playing with it.
Since many people call it the ChatGPT killer, let’s compare it horizontally with ChatGPT today.
According to the official statement, Claude’s strongest point is text creativity, so let’s test its writing ability first.
Here, Shichao first asked both models to write a horror story for me. The stories of the two players were well written, but it seems that the theme story is still a bit too simple for the AI large language model.
Let's do some more practical tests and let the model write some text that can really be used in work.
After all, the main purpose of Claude, which is connected to the work communication application, is to take it to work.
#First, let them directly provide some shopping guide words, and let ChatGPT recommend a 24k pure gold and diamond watch together with it.
Although, the copywriting given on both sides is very complete. But it is obvious that Claude, the shopping guide, has a smoother mouth.
He even said, "Because it is so rare, it is worth owning it today!" This whole group of people was a bit reluctant not to buy it.
Later, Shichao and the others fed me several popular title libraries and asked them to help me come up with some Xiaohongshu titles to promote ChatGPT.
The answers on both sides are good and can be considered a complete title. However, the World Super League would prefer Claude’s answer, which is a bit authentic Xiaohongshu. In comparison, ChatGPT's answer is a bit mediocre.
In this way, from the perspective of text capabilities, Claude’s text is indeed as official as it said, morecreativity.
The most important thing is that Shichao feels that Claude’s Chinese is quite authentic. Some words and expressions are particularly local.
In this round of Chinese writing competition, Claude is even better.
Let’s change a subject and try some math calculation questions.
First of all, let’s take a simple and classic chicken-rabbit problem. 10 heads, 28 legs, how many rabbits, how many chickens?
It seems that Claude is very good at calculations, setting up unknowns, and formulating equations. . When I looked at the answers, none of them were right. But ChatGPT gets it right easily.
Shichao changed the question to a simpler one and gave them a try. A frog in a 100-meter well jumps up 3 meters and falls 2 meters a day. When will he jump out? .
ChatGPT Three times five times two, two formulas to quickly get the answer.
But Claude counted five scraps of paper, and even after being guided by Shi Chao, he still didn't get it right. .
Although after being prompted, it realized that it would not slide another 2 meters on the last day, it still made a mistake in the second step. .
After all, even the officials themselves jumped out and said, we are not good at solving math problems, please feel free to ask. .
#In this round of mathematical competition, ChatGPT won without any pressure.
But it’s okay, it’s just a tie now. Let's take a look at the code to see if Claude can make a comeback.
Shichao tried to make them have a basic web page effect. In the blue web page, there is a white round button. When the mouse is moved over the button, it will enlarge into a square shape.
Both parties can understand Shichao’s needs and write appropriate code.
Claude’s web page effect:
##ChatGPT’s web page effect:
Even Claude thoughtfully helped me make rounded corners and added a slow-moving effect.
Seeing that it was performing pretty well, Shichao tried to add a number to it. Let it be a web version of Snake. But this time, what I did was not just the slightest bit. The greedy snake eats only one piece, and writes down the whole snake. .Snake game, where is the snake?
Looking at this point, it turns out that GPT-4 is still the big brother of AI. , you can make a complete Snake game in 20 seconds. The editorial department also tried to use GPT-4 to create a complete 2048 before. In terms of code, ChatGPT and Claude are evenly matched and tied.It seems that in the end, the two companies can only compete on the century problem of AI large model
.
Selected 100 questions from the Mentally Retarded Bar, arranged for two guests.Sure enough, when it comes to brain teasers, Claude and ChatGPT are evenly matched. .
The first question is, Xiao Ming’s father has three children. The first is called the eldest, the second is called the second, what should the third be called?
Claude's brain was short-circuited and he could never turn that corner. .
However, there are some questions where Claude will be better.
For example, "What was the name of the cobra before glasses were invented?" ChatGPT lost, but Claude could answer it perfectly.
After the above rounds of competition, ChatGPT and Claude were basically tied.
Claude’s text capabilities will be stronger and more in line with the usage habits of Chinese users. But in terms of math questions, it is much better than ChatGPT. . .
#However, I can’t stand it because it’s free, the threshold is low, and the Chinese experience is good.
Shichao recommends that all interested friends can go and experience it as soon as possible. There is no guarantee that Claude will start charging in the future.
Perhaps, when Anthropic is developed to the point where it can be charged, Claude will be able to compete with OpenAI.
The above is the detailed content of A former employee of OpenAI created a GPT killer, which is free to use and can be played with just one click.. For more information, please follow other related articles on the PHP Chinese website!