By Jenny Lee WIRED Korea
When the world’s greatest Go player Lee Sedol was facing off against AlphaGo in a Korean hotel room in 2016, more than 200 million people around the globe tuned in to see this historic match, which ended with a triumph for the Google machine.
What they were interested in was not simply seeing a machine beating the 18-time world champion at this enormously complex Chinese board game, but also getting a glimpse at how far artificial intelligence (AI) technology has progressed in terms of replicating the intuitive decision-making process humans have. It was clear many were baffled by AlphaGo’s ingenious moves.
While a sci-fi vision by the late British physicist Stephen Hawking of a conscious computer that is many times smarter than a human and can spell the end of the human race has not yet taken hold, AI has made some huge advances over the past few years, to the extent that it is now being put to practical use.
The dominant AI technique today is machine learning, which depends on large quantities of data to train neural networks to recognize patterns and make predictions, and this is ready to innovate and transform, if not already doing so, a number of industries including broadcasting and digital media.
Tech giants and broadcasting media companies are embracing this futuristic technology and infusing it throughout production workflows and video libraries. And, as noted in a 2018 report by the Electronics and Telecommunications Research Institute (ETRI) as well as a 2019 report by the International Telecommunication Union (ITU), the rewards of using AI in broadcast are significant – from increased efficiency and flexibility to cost savings during program production.
“Though it (the use of AI in broadcast) is still in the initial stage, there are many cases and projects currently underway across the industry” said Kim Song-min, a researcher at ETRI’s Intelligent Convergence Research Lab. “It is being tried and tested for extracting content from vast archives, automatically localizing content for international distribution and generating access services, such as captioning, audio description and text to speech, to name a few.”
Key Advantage: Automation
With broadcast workflows including a number of repetitive tasks and processes, much of the broadcast industry uptake seems to have been spurred by automation that AI technologies promise.
The BBC research and development team in 2018 tapped AI machine learning algorithms to delve into the treasures of the BBC archive. Computers trawled through thousands of hours of legacy content dating back to 1953, using information from past scheduling, content metadata and other program attributes, and generated programming across two full days, branded as “BBC 4.1.”
Lotte Home Shopping, a TV shopping channel of Korean retail giant Lotte, in 2018 also introduced the "Smart AI Programming System," which forecasts both the timing and the sales volumes of specific products and automatically creates a schedule that would maximize its sales.
“Metadata, or little content descriptors that help tell what each content is about, is important to have algorithms to work,” the ETRI researcher said. “Many broadcasting companies in Korea and abroad seem to be in the process of tagging content with rich metadata, a task for which they also turn to AI.”
In addition to automating and optimizing content programming, AI is also changing the face of video production.
For the first time in 2016, Watson, IBM’s AI technology, was programmed to find areas of high action or high emotion from the movie Morgan and make those selects to help an experienced editor create a trailer.
But now it can do much more than pulling selects. Watson’s machine learning algorithms can analyze scenes from a live sports event to put together effective montages of match highlights in a matter of seconds. The 2019 Wimbledon Championship, the oldest tennis tournament, was where this AI-driven technology shone.
Some other areas where automation has driven productivity and efficiency include speech recognition like the one developed by Japanese commercial broadcaster TV Asahi, which helps produce captions for live TV programs quickly and accurately and speech synthesis, or text-to-speech conversion, which enables broadcast programs to employ various styles of voices like those uttered by a CG-generated announcer at NHK Japan.
Tailored Audience Experience
Making use of AI technologies for more targeted, relevant programming that promises to improve audience experiences is also an attractive idea, which has taken a number of companies in the broadcast and media industries by storm.
A good example of this is Netflix’s unmatched recommendation system, in which algorithms analyze and detect patterns from data related to users' viewing habits to suggest the right content tailored to each of its users. On the platform, more than 75 percent of viewer activity is influenced by the recommendation algorithm, according to the U.S. streaming service which has 167 million subscribers worldwide.
This AI-driven recommendation engine results in a richer and personalized experience for consumers, which in turn leads to constant revenue generation for the brand.
One of Korea’s mobile carriers, KT, is also following the footsteps of Netflix by launching an over-the-top (OTT) service, Seezn, in November, which uses an AI-based content recommendation feature that recognizes facial expressions of users via smartphone cameras and analyzes their movements to suggest customized content that best matches their mood.
KT, in partnership with the other two Korean mobile operators SKT and LG Uplus, is now looking to provide relevant and interesting ads for each TV viewer, rather than those compatible for mass audience.
Addressable TV, a new advertising technique powered by machine learning, is what they are trying to develop and commercialize, and it works by analyzing household viewing histories and other data such as gender, age, characteristics and interests and sending each household relevant ads.
The technology will be incorporated into their Internet Protocol TV(IPTV) set-top boxes, which, according to the Ministry of Science and ICT, have penetrated into 16 million homes in Korea.
“While the three mobile carriers are locked in a fierce competition,” said Jo Jee-hyung, a researcher at Yonsei University’s Graduate School of Information, “they are cooperating on this front as the more data there is to analyze, the more sophisticated the algorithm becomes, and the more sophisticated the algorithm, the more relevant the match is.”
As of December, KT had a market share of more than 21.4 percent, followed by SK Broadband with 14.7 percent and LG Uplus with 12.4 percent, the Ministry of Science and ICT data shows.
Despite a number of AI applications in the broadcast industry, it is still years away until the technology comes of age, which suggests all the efforts that are being exerted at present may be just a snippet of what will be available tomorrow. That is one good reason to be excited about what the future may hold.
- 최초작성 2020.05.22 15:27
- 수정 2020.05.22 18:50