[ad_1]
Last December, the American media The New York Times first reported on OpenAI and Microsoft.Filing a lawsuitaccusing the other party of misappropriating millions of copyrighted articles of the newspaper. Artificial intelligence (AI) technology is advancing by leaps and bounds, but the controversy involving billions of dollars has not yet been resolved. Recently, the AI smart summary platform Arbor announced a cooperation agreement with the New York Times and local financial media “Hong Kong Economic Times Financial News” to launch the “Top Media Packages“.
The “Premium Media Package” is priced at HK$998 per year, or an average of HK$83 per month.Arbor AppAll paid content of the Hong Kong Economic Times can be viewed on the Hong Kong Economic Times website and app. As for the content of the New York Times, it can be read on its website and app.
Controversy surrounds cooperation between media, technology and enterprises
There has always been tension between the media industry and AI technology companies. Tech optimists argue that the AI trend is unstoppable, while traditionalists insist that words cannot be sold at a price. For example, the US News Corporation reached a “Multi-year global partnership“Agreement, authorizing the other party to obtain the content of well-known newspapers such as The Wall Street Journal. At that time, OpenAI can use these news contents to train the chat robot ChatGPT and enhance product capabilities.
As a “conservative” in the eyes of the public, The New York Times seems to have nothing to do with AI technology companies. We may use this collaboration to explore how media and technology companies should cooperate – a topic that has not been determined in both academia and industry.
Intelligently extract the essence of news
Different from using text content to improve the quality of chatbot answers, Arbor’s biggest selling point is information aggregation and smart summaries, which are suitable for Hong Kong people who want to get things done quickly. Ken, vice president of machine learning at Arbor, said in an interview that a piece of news is often reported by multiple media outlets, but the content is similar.
Arbor can, to some extent, “gather the best of hundreds of sources” and summarize lengthy news reports describing the same event into three sentences. It can also extract the key points of the article from the vast “sea of words” and provide a three-sentence summary for each key point. Finally, it will attach a link to the source of the news to facilitate readers to read the original text.
Talking about the amount of information on the Internet, Ken said with a smile, “The numbers are shocking.” The amount of data is “175 zettabytes (ZB), and 1ZB is equivalent to 1 trillion gigabytes (GB), which is actually a number that is so large that ordinary people can hardly imagine.” In this case, Arbor cannot be detailed. Therefore, for news events that are mentioned more times, the algorithm behind it will define them as high-value content and summarize them.
New video summary feature
Short video platforms such as TikTok are gaining popularity, showing that the public is increasingly demanding more concise content. Arbor recently launched a video summary feature that allows users to quickly access key information from a YouTube video by simply entering a link.
Ken admitted that summarizing a video is much more difficult than summarizing text. For example, in a personal interview, the video is usually mixed with a lot of content that is irrelevant to the main point, which is a kind of “noise” to the final result. On the contrary, the content of the interview presented in text is often to the point and accurate enough.
Self-developed model ensures accuracy
The most important thing for news reporting is fairness, authenticity and reliability. In the process of extracting the essence, how to avoid factual errors is a major challenge faced by the Arbor team. Ken pointed out that after the text is summarized by AI, it will also be tested by the hallucination detection model developed by the team.
Ken further explained, “The so-called hallucination is technically called actual inconsistency… Our team has built its own data set (for this purpose). Simply put, half of the summaries are correct and the other half are wrong… Based on this data set, we trained a (hallucination detection) model.”
After ensuring accuracy, the next step is customization. Arbor launched theNews Express“, providing readers with customized news insights. Users can let Arbor understand their reading interests through 10 questions. Ken said with a smile, “There are different roles in watching football, an ordinary fan or a professional gambler, looking at the same news from their perspective, the insights they expect to get are different.”
Arbor is an innovative AI information platform registered in the UK. It has recorded more than 100,000 users in about three months after its launch. Looking back on the entire R&D process, Ken admitted that he encountered many difficulties. How to deal with massive amounts of data, how to continuously upgrade model algorithms, etc., these are the “monsters” that the team has fought in order to “upgrade”. Ken said that how to accurately push news insights to the corresponding readers is one of the problems that “has not yet been solved in the market, but Arbor is working hard to overcome.”
Support EJ Tech
If you want to submit articles, report information, issue press releases or interview notices,Click here to contact us.
//
[ad_2]
Source link
