It appears that Artificial Intelligence (AI) research is getting a lot of attention in the US and the UK, but Japan may not be too far behind. In this blog, I summarize what is going on with AI research, products, and services in Japan. I need to warn you that some of the links are available only in Japanese. You need to take my word for it.
Although AI research, products, and services are not exclusive to large companies, the usual suspects are active in this area in Japan. Because AI covers a wide variety of technologies and domains, each company has its own unique focus.
My objective is to understand who the AI players are in Japan and to find out if there is any opportunity for US companies to collaborate or compete with them in the Japanese market. My unscientific method of skimming through a dozen or so articles in Japanese, including those in famous publications like Nikkei and IT Media, found a consensus that Japan needs to invest more actively in funds and human capital to catch up with the US and the UK.
The following is a list of companies I focus on:
In this blog, I cover only NEC; I’ll cover the rest in later blogs.
NEC issued a press release on November 11 this year about reinforcing its focus on AI, saying that it has been active in AI research since the 1980s. NEC claims it excels in voice recognition, video/visual recognition, natural language processing (NLP), machine learning (ML), prediction, and optimization/control. They plan to increase staff to 1,000 in the AI area by 2020. They are applying the results to creating a safe society and safe cities, monitoring the safety of infrastructure, predicting and controlling demand for resources, improving CRM, and optimizing business processes.
NEC focuses on the following three areas:
- Face recognition: NEC was number 1 in the Face Recognition Vendor Test (FRVT) by NIST.
- Super-resolution technology to make vague images clearer: NEC has a short video to introduce their technology.
- Crowd-behavior analysis to detect unusual movements and activities to avoid dangers and crimes. (NEC has a short video for this.)
- Object detection to tell the kind of materials composing an object, which is used to decide authenticity.
- Internal detection of a structure deterioration through video
- Failure Sign Monitoring System for Large-scale Plants
- Heterogeneous Mixture Learning to automatically discover massive patterns hidden in big data
- Implication clustering to grasp trends quantitatively by grouping text with the same meaning
- Text entailment recognition to recognize included relationships of semantic content between texts, the technology for rumor detection from cyber information and the semantic search technology for improving the operation efficiency of contact centers
- Rapid Machine Learning (deep learning) to produce trends by analyzing big data
- Heterogeneous Time Series Dynamics Profiling to search a visual target out of big data with face recognition.
- Autonomous and adaptive control technology
- Predictive Robust Optimization Framework Technology to make a decision using prediction
NEC has an assortment of AI technologies, as shown above. According to their press release, some of them are highly regarded and won prizes in NIST competition.
The technologies are meant to be used by system integrators to create a solution for each user. This approach is different from that of IBM Watson, which is to make their platform and APIs open to solicit wide participation.