Boost to university-industry AI research collaboration

As the global race in artificial intelligence (AI) research intensified over the past year, South Korea’s ministry of science and ICT recently announced a KRW2.2 trillion (US$2 billion) budget for AI research and development to build up talent and use its universities’ strong links with industry to become a global AI powerhouse by 2022.

“We aim to reach the global top four by 2022,” Chang Byung-gyu, head of the Presidential Fourth Industrial Revolution Committee – a body set up in October 2017 – said in May when the new AI funding was announced. The top three – the United States, Japan and China – are investing heavily in AI and related technologies and South Korea is leading a group of countries that aspire to compete with them, including Canada, Singapore, Taiwan and the United Kingdom.

“Artificial intelligence is not simply a new technology but a key driver of economic transformation,” Chang said, noting that globally the speed of AI development is accelerating.

“Korea is still behind the US and China in terms of AI technological competitiveness, investment and manpower. I think the sense of urgency toward this emerging technology pressured the government to make an investment plan,” said Sung-chul Shin, president of KAIST (Korea Advanced Institute of Science and Technology), in an interview with University World News.

The institute stands to benefit substantially from the increased funding, including for collaboration with industry where it has considerable experience.

“Korea is very strong in the fields of semiconductors, smartphones, electronic appliances and automobiles. When applied with advanced AI technologies, Korea’s flagship industries will be more dominant in the global market in the near future,” Shin said.

The funding announced in May will be used to set up six AI graduate institutions with the stated aim of turning out some 5,000 AI and data management specialists in the next decade and strengthening partnerships with industry in AI R&D.

According to Korea’s Institute for Information and Communications Technology Promotion, South Korea was third after the US and Japan in the number of AI-related patents issued from January 2005 to September 2017, and the rate of growth in patent applications and the number of quality applications were inferior to the US and China.

The ministry has described the country as being 1.8 years behind the US in AI research and adoption, and noted that in AI China surpassed Korea since 2016 – the year of the so-called ‘AlphaGo shock’ when the AI computer programme AlphaGo beat Korean grandmaster Lee Sedol at the strategic game of Go. Within days of the defeat, South Korea’s then president Park Geun-hye announced KRW1 trillion (US$884 million) in government investment in AI R&D until 2020.

Just three years into that programme, more than double that amount again has been pledged. The huge expansion in AI-related research funding around the world – but particularly in China where huge investment in AI was announced in the past year – meant Park’s 2016-20 programme was deemed insufficient to close the gap.

Talent cultivation

Under the latest plan, South Korea will cultivate 1,370 AI talents by 2022, including 350 key researchers, and will award 4,500 domestic AI scholarships. The ministry also announced a short-term project to address the AI talent shortage with six-month intensive training courses to “gestate 600 young talents” by 2021. Meanwhile, universities are being encouraged to set up AI courses.

Some 12 university ICT research centres would be set up at a dozen of its leading universities, each assigned its own research niche to train talent “in critical technologies of the fourth industrial revolution”, the ministry said. The research centres will receive KRW5.3 billion to KRW8 billion overall in research funding over the next four years.

Three of the research centres will be in blockchain research, AI and big data at Hoseo University, KAIST and Korea University. The research centre at KAIST will focus on deep learning for AI and will be involved in international joint research projects and public-private partnerships, not just with big corporations but also small and medium-sized enterprises, KAIST’s Shin said.

“We already launched strategic plans to expand AI education and research even before the government announcement in May,” said Shin. AI basic and advanced courses had been expanded from the current semester, he said, adding that “we plan to open the Graduate School of AI in 2020 to foster AI specialists”.

“We will establish the AI Institute next year,” he said. This will be for multidisciplinary ‘convergence R&D’ in the fields of ICT, unmanned vehicles, the Internet of Things, drones, robotics, medical services, new material design and information security.

“We put top priority on basic AI algorithm research but hope our technological applications will be embedded into high-tech electronic devices and other industrial sectors such as low power consumption, using less memory, and expanding computer power.”

Industry collaboration

Currently more than 150 KAIST professors are involved in AI research, according to Shin. But the institution is also known for its strong collaboration with industry. “KAIST has had a very long history of industrial collaboration,” he said. “I can say the success of [Korean companies] Samsung, SK Hynix, LG and Hyundai are the results of these collaborations over the last 40 years.”

In Samsung’s semiconductor division more than 1,000 research personnel are KAIST PhDs, said Shin, but more broadly, “one in every four PhDs in the semiconductor industry which is dominating the global market, is a KAIST alumnus”.

“Beyond the semiconductor area, we believe that AI will be the most import area of industry collaboration in the years ahead. We have a strong interest in research activities on AI semiconductor chips, including neuromorphic chips which will be the crucial components in mobile and IoT [Internet of Things] devices for AI processing,” Shin said.

The ongoing US-China trade war which has put a squeeze on semiconductor exports to China – which barely produces semiconductor chips domestically and is reliant on imports – is threatening to slow China’s advances in AI technologies and has highlighted the importance of advances in semiconductor technology.

Almost three-quarters of R&D investment in Korea is conducted by business and tends to be focused on manufacturing.

Overall, around a fifth of KAIST’s research funding comes from industry. The institution has long-running masters and doctoral programmes with Samsung and semiconductor giant SK Hynix. Over 100 professors work on projects with the two companies, Shin said.

Samsung is the leading industry collaborator in terms of funding and number of projects. “I think Samsung understands the importance of long-term R&D and education. And it attests to the success of Samsung in semiconductors and smartphones. We hope such a success story will continue in AI as well,” said Shin, who adds that the institution and Samsung both “understand the dynamics of mutual collaboration”.

Not just big corporations

“Usually, industries tend to have short-term results. The most important approach [for universities] for avoiding pitfalls [in collaborating with industry] is to design and execute long-term research and education programmes. And the research themes should be fundamental issues so that universities and industry can learn and benefit together,” Shin said.

But collaborations with Samsung have not always gone smoothly.

In June Samsung Electronics was reportedly ordered to pay US$400 million to KAIST’s licensing arm after a US federal jury in Texas ruled the company had infringed its semiconductor-related technology patent FinFet – a transistor technology used for mobile phone processors.

Samsung refuted that it worked with the Korean university to develop the technology, denied patent infringement and challenged the validity of the patent during the trial. Shin declined to comment on the case.

Advances are not just made by big conglomerates like Samsung, which in August announced a US$22 billion investment plan for AI over the next three years and that it would hire AI researchers at some 1,000 AI centres around the world. Start-ups and smaller, more agile companies have also become crucial in the AI race.

“The collaboration model with industry varies,” Shin said. “For instance, we opened more recently the AI Quantum Computing Information Technology Research Centre funded by the government and four companies. Funding came from [Korean telecoms giant] KT and three small and medium-sized tech companies in Korea.”

In addition, one of KAIST’s tech start-ups was involved in the consortium. “We are trying to build an R&D ecosystem,” Shin said.