When it comes to AI in Japan, Masked Analyze is well known in the industry. His experience with both AI startups and large enterprises have earned him a wealth of connections, and since starting his twitter account (@maskedanl) in September of 2017 he’s amassed a large following for his updates on AI-related news. He regularly attends and presents at AI events such as the AI Business Innovation Forum organized by FRONTEO, and writes for a variety of IT media.
Masked Analyze has a deep understanding of the AI industry thanks to his link to data scientists and entrepreneurs on twitter, and his support of a range of AI-based startups. He’s also constantly on top of the shifting landscape of technological implementation, company press releases, and twitter developments,
In this interview, we ask Masked Analyze for his thoughts on the most recent trends in the Japanese AI industry and what they mean for the future.
Current trends and issues for AI in Japan?
Masked Analyze: There’s still a persistent worry around the idea that robots will steal human jobs. But even if the career market and jobs change, I don’t think this means that jobs themselves will disappear. Recently AI in Japan is also seeing implementation in games and smartphone applications. Because there are also companies that haven’t even started on AI implementation, I think there’s lots to look forward to in terms of digital transformation in the future.
There’s also an increase in the application of AI and robotic process automation (RPA). RPA is a good fit for the current state of work in Japan, where job responsibilities often include large amounts of straightforward administrative tasks. RPA is a solution that could bring constant results with regards to Japan’s declining work force and aging population.
Successful AI implementation and failed AI implementation: What’s the difference?
Masked Analyze: Excellent data scientists put their focus on developing models with high levels of accuracy, but that alone won’t result in the successful implementation of an AI project. Firstly you need to have a concrete problem that is AI-solvable, and to set realistic KPIs for the project. There’s also the matter of data challenges at the preparation level; if your company doesn’t have a sufficient amount of appropriate data for your project, you need to have plans in place to collect new data as necessary.
Projects will fail when the mindset is one of “I’ll find data scientists and they’ll take care of the rest.” You want to avoid jumping straight into large scale projects without any knowledge and experience. Lack of communication can result in data scientists building an AI model that does not meet your actual needs in the field. This invites failure. I recommend starting with a small proof of concept, building knowledge and experience, and expanding the project from there.
Also, after you implement a machine learning model, your AI project still isn’t done. A company’s goal isn’t “implement AI”, but rather “implement AI that results in improved work efficiency.” So you also need to be reviewing data to maintain an accurate model while analyzing cost vs. effect.
Big Data and Protecting Individual Privacy
How do we protect privacy in the age of AI?
Masked Analyze: On the service user level, people need to check their user agreements and get an understanding of how services and data are connected. It’s hard to imagine a world without cloud services and applications, so it’s important to choose services you can trust.
On the service provider side, you have to ensure security, but a perfect solution is very difficult. Outside of large scale individual information leaks, the news media doesn’t cover privacy extensively. And though Japan has a domestic data center, that doesn’t mean we can affirmatively assert that it’s safe. I want to see more focus on improving law that includes the protection of individual data. It’s not just privacy leaks, but also how an individual’s data is used; regulation is important. I also hope to see data scientists use open data for more research, and regulations that make it easier to access data for research purposes.
Sometimes Japanese companies are concerned with security to a fault. I’d like to see an equal level of focus on productivity and innovation, along with security. Data use and information security differ depending on the country, and use of data and AI in Japan will be impacted by what policies the country decides upon.
Challenges for Japanese enterprises in the age of AI
What should Japanese companies focus on to survive?
Masked Analyze: It’s difficult for Japan to go head to head with America and China, who are at the forefront of research and development. For Japan, victory lies in continuing along the lines of how to implement AI to benefit society, as well as specialization in specific areas, such as manufacturing.
As AI startups increase in number, I want to see a phase where we weed out the companies with technological know-how and the ability to implement AI from the companies without. What we’re seeing now is a failure of AI implementation because the companies making the tech don’t have the skills, so I think high levels of literacy will be important on the side of those hiring AI implementation services.
I think as programming becomes a part of compulsory education in 2020, we’ll see new opportunities and abilities as a result of kids being in contact with programs. Most adults now approach programming with some apprehension, but I hope that by being in contact with computers and learning how they work, kids won’t have this same apprehension. Even if you don’t like programming, just using computers will lead to higher levels of IT and AI literacy. I think this is going to impact all curriculums, across teachers, languages, and teaching methodology.
And I know this is self-promotion, but I talk about this and IT literacy in further detail in my book “Korekara Data Science Business (これからのデータサイエンスビジネス）” (in Japanese), which is co-authored by Kentaro Matsumoto. It’s aimed at business people in their 20s to 40s who work with information technology as part of their company.
Lionbridge provides high-quality data for machine learning at scale. With over 20 years of experience as a trusted training data source, Lionbridge AI helps businesses large and small build, test and improve machine learning models. Our community of qualified contributors is located across the globe and available 24/7, providing access to a huge volume of data across all languages and file types.
Looking to build or train a machine learning model? Get in touch.