BreakThrough Interview

―Understand AI & Experience Real Results―
AI Introduction Support Program (Machine Learning & Deep Learning)SOUYA Co.,Ltd

2018.10.30

SHARE
  • facebook
  • twitter

AI solutions for your worries & questions

summary

  • Souya offers an “Introduction Support Program” for companies who want a trial experience of the results of AI.
  • When supporting each customer’s unique worries and questions, the key to successful implementation is the right collection of datasets.

The news seems to be full of stories about artificial intelligence (AI). However, most people can’t quite understand how AI works, or imagine how AI could benefit their workplace.
Souya Co., Ltd. developed its “AI Introduction Support Program (Machine Learning & Deep Learning)” to address the worries and questions of those companies. Souya previously focused on app and software development - the “sou” of Souya comes from the Japanese word for “create.” President Takuya Matsumura explains that this name was chosen to reflect the company’s philosophy, “We want to continue thinking creatively.”

Many companies are reluctant to introduce AI into factory floors due to start-up costs and time. Souya’s projects allow clients to verify the results of AI introduction in a multi-step process to alleviate concerns, answer questions, and promote the understanding and normalization of AI. We talked with Matsumura about the services his company offers.

What exactly is “artificial intelligence”?

Simply put, artificial intelligence (AI) is an artificial system, usually a computer, that performs intelligent tasks and makes decisions at a level that was previously only possible for humans. New high-level technologies are constantly emerging in this field, and AI is already implemented in our daily lives in the forms of search engines, house cleaning robots, and self-driving cars.
Manufacturing industries are watching AI tech eagerly. If they can replace part of their human workload with AI, they can expect to avoid labor shortages and improve productivity while reducing costs. Large corporations are already implementing AI into their workforce.

The Core of AI Tech: Machine Learning & Deep Learning

The most important technologies behind AI are “machine learning” and “deep learning.” Machine learning refers to a computer reviewing repetitions of huge amounts of data to identify and learn patterns and rules to provide a reason or answer to other separate tasks. Deep Learning technology is a hot topic right now. It uses “neural networks” similar to a human brain to perform highly precise reasoning.

For example, a manufacturing site could use cameras and AI to substitute for workers performing visual inspection to identify defective products. In an office, AI could be used to categorize documents by scanning diagrams and cross-checking for similar data in its database, and searching for and identifying important keywords. This time-consuming task could be handled by AI to improve work efficiency and reduce workload.
Other recent applications include generation of CAD data from blue prints, automated identification and masking of human faces in photos and video, creation of diagrams and editing of photos and video.

Souya’s AI implementation (Example 1): Automated document summary using keyword scanning to improve office efficiency
Souya’s AI implementation (Example 2): Detection of defects using sensor data in industrial fields like factories

The key to successful AI implementation is a plentiful collection of datasets

Matsumura explains, “Some people think that AI implies superior intelligence, but it’s not always easy to get satisfactory results.”
The core of AI tech is “learning.” In order for AI to be useful, it must start by learning from images, audio, or text data. The data used for the learning stage is called “training data.”
Every successful implementation of AI demands thorough preparation of training data for efficient learning. Proper preparation of training data is the most laborious step of introducing AI.

For example, a system to detect defective products must begin with collecting as much data about defects as possible. Next, teaching the machine how to distinguish between good and defective products, and categorize by different types of defects, before the system can be used for inspection.
In other words, even if a company implements AI to improve the efficiency of defect detection, it will not learn properly if the training data is insufficient. In cases like this, Souya may recommend to simplify system output, like changing the method of approach from “detect defects” to “identify good items” (something that doesn’t require additional training).

AI implementation may sound simple, but many companies don’t realize that it requires preparation. Companies may not have a collection of defect data, or the data may not be in a format that is usable for system training. For companies who are unprepared, Souya offers data collection and accumulation support. Because the organization, cataloging and correction of data is a heavy burden on a busy client’s workload, Souya’s support must begin at this essential stage to allow for smooth AI introduction. Souya is confident in its data gathering capabilities and processing know-how.

Experience Results From a Small Start

In order to introduce AI successfully, it is necessary to first set a clear and firm goal of where and how it will be utilized. Matsumura explains, “First, we want our clients to see the potential of AI in the most obvious data and processes.”
The introduction support program begins by addressing a small task, placing importance on producing verifiable results from AI in a short time. Clients learn about AI and observe concrete results, allowing them to see the value in the results compared to the cost.

The consulting stage uses real data for a preliminary investigation, and analysis is performed at multiple stages of the process. If the client agrees to a contract based on those small results, the planning content and other costs will be revealed. System development, demonstration, verification and payment will occur after system introduction is agreed upon

Will AI replace human jobs?

Clients often say that they want to reduce labor costs by substituting AI for simple human labor. AI’s strongest point is that unlike humans, it is never affected by exhaustion, its performance is stable and predictable. Furthermore, AI has special capabilities that humans cannot compete with, like processing massive amounts of data.

But, if you ask “Will AI replace human jobs?” Matsumura will always say “NO.”
“AI works together with humans. The use of AI allows humans to refine their natural strengths. When AI provides a hypothesis, when it is focused on achieving a goal, when it is facing a potential problem that people can feel... all of this is only possible with the cooperation of humans. The more I work with AI, the more potential I recognize in humans. If we can think of artificial intelligence as expanded intelligence... humans and AI will work together, we will make use of each other, and we will never see the end its potential.”
Souya is actively and ambitiously involved in collaborative research in highly original fields, including work utilizing AI for high-speed calculation. They are advancing AI support for their client companies and searching for fresh ides to pull dreams out of fantasy and into reality.

Interview Date: September 25, 2018


Souya Co., Ltd.

For 13 years, Souya developed business apps for a number of industries. Fusing that experienced business knowledge with AI (machine learning and deep learning) and data analysis enables Souya to introduce AI systems that give client companies an immediate advantage.

share by SNS

related article