Reddit to pay companies like OpenAI to avoid inappropriate use of user chats

News on April 19th, Microsoft, Google and OpenAI have been using user chat data on Reddit forums to develop new artificial intelligence systems. Now, Reddit plans to charge these companies royalties.

Reddit has always been a hub for discussion of hot topics on the Internet, with approximately 57 million people visiting the site every day to discuss everything from makeup tips to video games to automatic car washes.

In recent years, chats on Reddit forums have become free training tools for companies such as Google, OpenAI and Microsoft to develop artificial intelligence systems. Now, many in the industry believe these artificial intelligence systems will be the next big thing in the tech industry.

Therefore, Reddit intends to charge companies that develop artificial intelligence technology. Many companies download and process forum chat through application programming interfaces (APIs). On Tuesday, Reddit said it planned to start charging companies to use its API.

Reddit founder and CEO Steve Huffman (Steve Huffman) said in an interview: "Reddit's database is indeed valuable, and we do not need to provide this value for free to the world's largest Company."

Reddit's move is the first time a social network has explicitly charged OpenAI and others for open access to develop artificial intelligence systems like ChatGPT. New AI systems like ChatGPT may one day become big business, but they won't do much to help companies like Reddit. On the contrary, these artificial intelligence systems can also automatically generate chat content and become competitors of Reddit.

Reddit is preparing for a possible listing this year. The company was founded in 2005 and currently relies mainly on platform advertising and e-commerce transactions to make profits. Reddit said it is finalizing the charging details of the API interface and will announce the price in the next few weeks.

Today, large-scale language models have become an important part of developing new artificial intelligence technologies, and user chat content on Reddit forums has become a valuable commodity.

One of the underlying algorithms of Google’s artificial intelligence chatbot Bard was trained using Reddit chat data. At the same time, OpenAI’s ChatGPT also uses Reddit data as one of the information for training large language models.

In addition, other companies are beginning to realize the value of chat content and images stored on the platform. Image hosting service Shutterstock has sold image data to OpenAI, which helped develop DALL-E, an artificial intelligence system that can generate images based on simple text prompts.

Currently, thousands of companies and developers, large and small, are using APIs to track millions of chats on the Twitter platform. Last month, Elon Musk, the owner of the personal social media platform Twitter, said he was changing the current way of using Twitter's API, charging fees ranging from tens to hundreds of thousands to use the API. But Musk didn't mention large language models as the reason for the change.

In order to continuously improve models, artificial intelligence companies need two important factors: powerful computing power and a large amount of available data. Some large artificial intelligence development companies usually have enough computing power, but still look for the data they need to improve their algorithms on the Internet. These include resources such as Wikipedia, various digitized books, academic articles, and chats on Reddit forums.

Companies such as Google, OpenAI and Microsoft have not yet responded to Reddit’s plan to charge.

For a long time, Reddit has had a co-dependent relationship with search engines like Google and Bing. They automatically obtain Reddit page information, index it, and then display relevant information in search results pages. Although this automatic crawling method may not be popular with all websites, Reddit can rank high in search results.

The large language model is completely different. It needs to obtain as much data as possible so that a new artificial intelligence system can be created.

Reddit believes that its forum data is particularly valuable because it is constantly updated. This freshness and relevance are exactly what large language model algorithms need to produce the best results, Huffman said.

"Reddit is better for chatting than anywhere else on the Internet," Huffman said. "There's a lot of content on the site that you would only say privately, or not say at all."

Hoffman also stressed that for developers who want to develop applications that help people use Reddit For now, the API is still free. For example, developers can use tools such as APIs for free to develop robots that automatically track whether user comments comply with content publishing rules. People studying Reddit data for academic research or non-commercial purposes will also continue to have free access to the data.

Reddit also hopes to integrate more machine learning into forum operations. For example, Reddit can use machine learning to identify the use of artificial intelligence-generated text on the platform and add labels to inform users which comments are from robots. . Reddit also promised to improve software tools for forum moderators to help them monitor third-party bots on their forums.

But for AI makers, Reddit thinks it’s time to pay up.

“It’s our own problem to create value by scraping Reddit’s data and not reimbursing our users,” Huffman said. "Now is a good time for us to strengthen management." (Chenchen)

The above is the detailed content of Reddit to pay companies like OpenAI to avoid inappropriate use of user chats. For more information, please follow other related articles on the PHP Chinese website!