IBM leads Microsoft, Toshiba, and Samsung in AI-related patents: UN

Big Blue has 8,290 patented AI inventions to its name, with Redmond owning 5,930 of them too.

Since the emergence of artificial intelligence (AI) during the 1950s, nearly 340,000 AI-related inventions have been filed for patents, a report from the UN World Intellectual Property Organisation (WIPO) has revealed.  

According to the WIPO Technology Trends report [PDF], IBM currently has the largest portfolio of AI patent applications, with 8,290 patented inventions. This is followed by Microsoft, who has 5,930 inventions.

Toshiba, Samsung, and NEC -- who possess the third, fourth, and fifth largest portfolios -- have 5,223, 5,102, and 4,406 AI patent applications respectively.

Despite IBM and Microsoft having the biggest portfolios, the Chinese state-owned SGCC has enjoyed the greatest growth from 2013 to 2016 in amount of patents filed, with a remarkable 70 percent annual average growth rate. Most companies have also shown increases in filing activity since 2012, the report said. 

In this respect, AI is currently undergoing a renaissance of sorts, with over half of all patenting activity in AI having taken place since 2013, which amounts to around 170,000 different patented ideas.

The ratio of scientific papers to inventions has also decreased from 8:1 in 2010, to 3:1 in 2016, indicating that AI technologies are being used more prevalently in the commercial sphere.

"Patenting activity in the artificial intelligence realm is rising at a rapid pace, meaning we can expect a very significant number of new AI-based products, applications, and techniques that will alter our daily lives" WIPO Director General Francis Gurry said.

SEE: How to implement AI and machine learning (ZDNet special report) | Download the report as a PDF (TechRepublic)  

However, there are a few companies that have decreased such activity. This is a reflection of companies using different strategies, WIPO said, with filings by Alphabet declining due to it having acquired 18 AI companies since 2009.

Other companies that have been active in acquiring other AI entities include Apple and Microsoft, with 11 and nine AI-related acquisitions apiece. 434 companies in the AI sector have been acquired since 1998, and 53 percent of acquisitions have taken place since 2016.

Of the organisations applying for patents, companies represent 26 out of the top 30, with the remaining four being universities or public research organisations. From the top 20 companies that filed the most AI-related patents, 12 are based in Japan, three are in the US, and two are in China.

Machine learning is the most used AI technique in patents, being included in 134,777 patent documents, which accounts for more than one third of all identified inventions. Filings of machine learning-related patents have grown rapidly, with an annual average growth rate of 28 percent between 2013 to 2016. The amount of machine learning-related patents filed has also more than doubled, rising to 20,195 patent applications filed in 2016 from 9,567 in 2013.

The key machine learning techniques being developed, the report showed, are deep learning and neural networks: Deep learning patent filings increased at an average annual growth rate of 175 percent from 2013 to 2016, with 2,399 patent filings in 2016; and neural networks grew 46 percent over the same period, with 6,506 patent filings in 2016.

Among AI functional applications, computer vision is the most filed type of patent. Computer vision applications include those related to augmented reality, biometrics, image and video segmentation, character recognition, object tracking, and scene understanding. Computer vision is mentioned in 49 percent of all AI-related patents, amounting to 167,038 patent documents. Japanese and Korean companies active in consumer electronics, imaging, telephony, and software, such as Toshiba, Samsung, Canon, Fujitsu, and NEC dominate the largest functional application, computer vision, according to the report.

While AI-related technologies can be applied across various industries, transportation was the prominent industry in for AI-related patents filed, growing by 33 percent annually from 2013 to 2016, and reaching 8,764 patents filed in 2016.  

AI-related telecommunications also experienced strong growth, with the amount of patents filed rising by an average of 23 percent between 2013 and 2016. 6,684 filings were made in 2016 for AI-related telecommunications.

SEE: The next step for machine learning and AI  

China and the US are the two most popular offices for filing AI patents, followed by Japan. Together, these three offices account for 78 percent of total patent filings.

The dramatic increase of patents filed in China reflects its increasingly important role in the tech sphere.

Despite China ranking first in the number of patents filed, only four percent of patents first filed in China are subsequently filed in other jurisdictions, compared to the 25 to 63 percent rate in all other offices. The high percentage of applications filed only in China could be due to Chinese applicants being more interested in the domestic rather than overseas market, WIPO said.

In the ongoing trade war between China and the United States, the topic of technology ownership has been at the eye of the storm, with the United States earlier this week unsealing a pair of indictments against Huawei.

The first indictment alleges that Huawei conspired to steal intellectual property from T-Mobile and subsequently obstructed justice. The alleged activity occurred during 2012-13, and relates to Huawei's attempt to build a robot similar to the one T-Mobile was using at the time to test mobile phones.  

Huawei is also facing a second indictment for conspiracy to defraud the United States; violate the International Emergency Economic Powers Act (IEEPA); violate the IEEPA; commit money laundering; and obstruct justice. 

Looking ahead, WIPO in its report also emphasises the need to address challenges surrounding AI technology, especially in protecting data privacy and security.

Currently, there are no standards for the anonymisation and sharing of insights from big data, Chief Data Scientist of UN Global Pulse, Miguel Luengo-Oroz said.

"New frameworks are needed that go beyond privacy and also ensure accountability and the responsible use and re-use of data for the public good."

Multinational corporations such as Google, Intel, Microsoft, and IBM are also releasing their own set of principles for developing responsible AI, which poses risks and future unknowns due to a lack of transparency towards citizens, according to the report.

"Citizens, private industry, national governments, and international institutions must work together to collectively establish standards for training datasets and auditing procedures that require the perspectives of diverse disciplines and coalitions," said UN University Center for Policy Research Fellow, Eleonore Pauwels.

Related Coverage

China's AI scientists teach a neural net to train itself

Researchers at China's Sun Yat-Sen University, improved upon their own attempt to get a computer to discern human poses in images by adding a bit of self-supervised training.

This is what AI looks like (as sketched by AI)

The self-portrait raises questions about the nature of creativity and human expression.

AI Edge X: The first 4G Industrial Gateway for AI on the Edge

Bridging the gap between the edge and the cloud in order to bring accelerated vision features and AI performance to the edge of the Internet of Things.

IBM's AI predictions: Trusted AI, quantum computing take center stage in 2019 (TechRepublic)

Major breakthroughs will lead enterprises to increasingly advance, scale, and trust artificial intelligence, according to IBM.