She wrote an article on ChatGPT and earned $14,000 a day

Write a blog with ChatGPT, and you can achieve a daily income of more than 10,000!

No kidding, it really happened.

The cause of the incident was that a young lady was preparing to publish a data analysis article, which had a word count of about 3,000.

But she felt that it was too troublesome to sort out the text content while analyzing the data, so she moved and asked ChatGPT for help.

At first, she didn't expect much from the results, but when the profit figures came out, she directly shocked me:

With a daily income of $2,000 (about 14,000 yuan), it's hard to believe that an article written by AI has such an impact!

Not only that, according to the description of the young lady, this article was also bombarded with Google-related content search No. 1.

While many netizens expressed amazement at this result, they also thought that this was a very good example of "human-machine combination".

How are AI articles with more than 10,000 daily incomes refined?

By chance, the young lady discovered a set of tables with company data on the Airtable website. (Airtable can be thought of as a kind of cloud Excel.) ）

Then she felt that if she could use that data to write a blog, it would be much stronger than cold numbers.

But problems followed, although Google can bring down these data through crawlers, but the work effect on data ranking and other work is not ideal.

This affected her analysis and refinement of data in the process of writing; Moreover, cramming so much data into a 3,000-word article was also a headache for her.

So, the little sister thought of the popular fried chicken in the AI circle - ChatGPT.

First, she saved her Airtable form as a PDF and uploaded it to Google Docs.

Then the little sister copied the table in Google Docs, pasted it into the ChatGPT dialog box (pasting a small part at a time), and gave the command:

Based on this data, write a paragraph.

But the little sister did not directly "take the ism", but when ChatGPT entered the paragraph, she also carefully checked the content for errors and edited the text accordingly.

Then she asked ChatGPT to title the article and write a summary; I found some copyrighted pictures and stuffed them into the article.

Finally, click "Push" and you're done.

As for the result, as just mentioned, this article made the little lady make $2,000 overnight. But what she didn't expect was that some of the traffic came from Google!

As early as last year, Google took action to treat AI-generated content as "spam" during the search process, resulting in a downgrade in the search ranking of the corresponding site.

This is as if opening a new continent for the young lady.

Because of the work she does, it is through the niche site (niche site) to do SEO (search engine optimization).

The niche site is similar to the domestic website version of Taobao: help Taobao sellers promote products and get commissions according to the transaction effect.

Its profit model is to obtain search traffic from Google, direct customers to e-commerce such as Amazon to place orders through high-quality content, and then share from it.

At this time, there are friends to ask, this article of the young lady, what is the proportion of "people" and "machines"?

The young lady directly announced the answer to this:

The proportion of AI-generated content is about 40%.

AI writing is caught in a storm of infringement

Then again, it's not for nothing that ChatGPT writes articles "so well and fast."

Just like the previous explosion of Stable Diffusion and DALL· Just as models such as E "borrow" the works of many masters when generating images, ChatGPT also "references" the works of peers when writing articles.

According to Bloomberg, in order to know which websites ChatGPT will refer to when "writing", a brother from the Wall Street Journal (WSJ) also specially asked this figure for verification:

What specific media content sources has ChatGPT trained in? Lists the most important content sources in the database.

ChatGPT did not hide here, and directly listed 20 media organizations:

Reuters, The New York Times, the BBC, Bloomberg and other big media are prominently listed.

At this time, many media and media practitioners can't sit still (after all, it also affects their own jobs):

OpenAI uses our articles to train ChatGPT, but doesn't pay for it???

Even Jason Conti, general counsel at Dow Jones Publishing, released a statement directly:

Dow Jones does not have an agreement with OpenAI, and if you want to use Wall Street Journal work to train AI, you must obtain a license from Dow Jones.

We take the misuse of our work seriously and are reviewing the situation.

At the same time, CNN also believes that training ChatGPT with its articles violates its terms of service, and CNN's current plan is to approach OpenAI and ask the latter to pay for content licensing.

Although most of the caliber on the media side is that "scraping data without permission will violate the publisher's terms of service", some netizens still raised different opinions:

AI's Robots .txt detailed the crawling strategy, and no protocol is required to crawl websites. Unless you have more solid evidence, this looks like a speculative accusation of not understanding how the network works.

However, WSJ brother did not show weakness, and directly moved the terms of use of the Associated Press, which roughly means:

It may not be authorized when it is not used for commercial purposes;

The intellectual property rights of the content, etc. remain here at the Associated Press;

not for any commercial purpose;

The Media reserves the right to remove references to the other party at any time without reason.

Of course, the "infringement problem" seems to be one that all AI faces.

At the end of January this year, Getty Images sued Stability AI in the High Court of London for copyright infringement.

Copilot, an AI programming tool, was sued by programmers for $9 billion.

And now that it's ChatGPT's turn, how will it weather its infringement turmoil?

One More Thing

Using AI to write an article, in fact, Qubit did it once last week - "Why ChatGPT is so strong: a detailed explanation of the 10,000-word long article by the father of WolframAlpha".

And this article is currently collected by 1300 users on Zhihu, and even some netizens said:

This is probably the clearest article I've seen about Transformer and GPT language models.

Reference Links:

[1] https://twitter.com/NicheSiteLady/status/1626576195688009734

[2] https://twitter.com/NicheSiteLady

[3] https://www.bloomberg.com/news/articles/2023-02-17/openai-is-faulted-by-media-for-using-articles-to-train-chatgpt?srnd=technology-vp

Jin Lei Pine comes from the Temple of Cave Fei

Qubits | Official account QbitAI