As hurdles are lower than before, more and more cases are developing artificial intelligence in-house. Data is the key to AI development. And, ahuman resources become indispensable for the data to be collected and created appropriately. What are the data issues, and what are The AI human resources who solve them? Mr. Satohiko Ishikawa, Ceo of Idemy, and Cedric Wagrez, Director of Lionbridge Japan AI Division, met.
The democratization of AI is an urgent need to “nurture” human resources.
Wagrez: Lionbridge began its AI business in 2007 and launched Lionbridge AI in Japan in 2019. According to the project requested, approximately one million workers from around the world registered at Lionbridge collect and create data, annotate, and evaluate AI models. As i develop my business, I have been feeling the democratization of AI for several years. I think that the hurdles to AI development have dropped and more companies have started to work on it, but how do you see it?
Mr. Ishikawa: Recently, the essence of the competitive advantage of companies has changed, with the digital transformation of DX(AI) as the basis for the change. At Aimee, we often support manufacturing companies, but we are shifting to the perception that hardware manufacturers will not be able to produce good products if they don’t make software. A prime example of this is Apple’s iPhone, and tesla’s electric car is recently. When that happens, there are many companies that are working to develop AI human resources because they cannot make good manufacturing without people who have a professional or literacy in the company.
Wagrez: Especially in Japan, there is a shortage of AI personnel, making it difficult to hire. That’s why it’s so important to bring up the people we’re with, and we’re focusing on The Service of Aimee.
Mr. Ishikawa: The concept of Aidemy Business, which we offer, is “In-house manufacturing support for AI systems.”. In addition to supporting the development of AI personnel through an e-learning service called Aidemy Business Cloud, we also offer a platform called modeloy moderoy. Modeloy aims to reduce software engineering effort by providing maintenance and management systems, including IoT collaboration that occurs after the machine learning model is created, and management applications required in conjunction with AI.
It doesn’t necessarily matter if you have the data you need.
Wagrez: Around 2007, when we started our AI business, we were still rule-based, but around 2012, deep learning technology came into play. The more data you have, the better the performance of deep learning, and the more companies are investing in data creation. What are the data challenges in developing AI?
Mr. Ishikawa: As data is said to be oil in the 21st century, companies are well aware of the importance of data. Perhaps because of this, we often start projects from the perspective of whether we can create data at hand and do something. It’s not a bad thing to think about data as a starting point, but it usually doesn’t work. This is because the business impact perspective is neglected with data.
Wagrez: Trying to develop AI without clarifying the problem is a solution looking for a problem, but as Ishikawa says, it is very important to clarify the KPIs that should be improved and to plan how to improve them. First of all, it is necessary to think about what data should be collected and created to solve the problem.
Mr. Ishikawa: I was very sympathetic to Wagrez’s “Solution looking for a problem”, so I’m going to use that expression from now on. That’s right, and it’s still the business impact of solving the problem, and it doesn’t necessarily matter that we have the data we need at the moment. If the cost savings of 1 billion yen per year are effective, the option of collecting data over 100 million yen will also arise. If it is difficult to complete in-house, there is also a method of outsourcing to external services such as Lionbridge AI, which collects and annotates data.