When it comes to implementing AI solutions in an industrial environment, we have faced various challenges. From my perspective, the main challenge is the data challenge, as we often have a large amount of data that is sparse and highly repeated in parameter space, as opposed to dense data. Additionally, in the production industry, there is a vast amount of know-how that is stored in theory, formulas, or people’s heads, which makes it essential to find a way to combine this know-how with sparse data. To address this issue, I like to approach the ssituation using a Quadrant matrix, which considers whether the data is sparse or dense and whether the know-how is rich or poor. This helps us determine the correct approach to take for specific tasks.
Moreover, in SMEs, it is crucial to take an agile, full-stacking approach, considering all aspects of the project, such as physics, engineering problems, applications, infrastructure, hardware, software, AI model, computation efficiency, accuracy, optimization, user experience, server, development, operation & maintenance, fast prototyping, communication, and project management. It is undoubtedly challenging but also exciting to address all of these factors simultaneously.
At the end of the day, AI is the key to stimulating production from intranet to collaborative production, and we are committed to finding solutions to these challenges.