Net Smile’s development flow is quite different from that of normal contracting-style system development.
Doing this the usual way, the process begins with defining requirements, after which detailed meetings are required during the entire process until full-scale operations begin. However, at Net Smile we use a POC (Proof of Concept) process that allows us to create prototypes and start running immediately after the concept planning meeting, allowing deep learning to begin at an early stage.
At Net Smile, rather than pursuing 100% accuracy from the beginning, we introduce a human-like AI (more than 99% accurate) that factors in mistakes of about the same degree as what would arise artificially (about 1%). By "training" the AI, we can improve accuracy and reduce total cost. AI can attain higher levels of accuracy after being trained via deep learning than before being trained. This training is a vital part of putting AI to practical use.
To develop AI, comprehension of technical English and advanced programming skills are essential.According to the Patent Office' s “2014 Report on Trends in Patent Application Technology,” only 2% of papers on AI technology are in Japanese. In order to understand foreign papers and the latest trends, a high level of English comprehension is necessary.
In addition, we recruit personnel who rank S (top 2%) in a programming skills check. Our strength is that we have a great variety of staff with high-level English comprehension and advanced programming skills.