Software Development in the AI Era (Part 2): Creating Design Assets Before Letting AI Write Code
ovitWriting code with AI will likely become the norm in the future.
However, from my perspective as someone who has been developing visual inspection software for over 20 years, it’s not the code itself that’s truly valuable.
It’s the experience I’ve gained on-site, asking myself, “Why was it designed that way?”
Source code is both an asset and a liability.
We have been developing visual inspection software for over 20 years.
As a result, there is a massive amount of source code.
Of course, some of those implementations are now outdated.
At the same time, it’s packed with know-how that has been refined through trial and error at countless job sites.
For example,
- Why is the processing order set up this way?
- Why is this exception handling necessary?
- Why Did You Choose This Algorithm?
- Why communicate with the equipment at this particular time?
These reasons are not easy to understand just by reading the source code.
Source code tells you " how it was implemented," but it doesn't explain " why it was designed that way."
What I want to pass on to AI is not code, but design philosophy.
So, what we're currently working on is the task of extracting design principles from existing software.
For example,
- Image Acquisition
- Positioning
- Defect Detection
- judgment
- Results Management
- Equipment Control
Features such as these, rather than simply as programs,
"What is the purpose of this feature?"
We organize the information to that level and store it as Markdown-formatted documents.
As a result, the AI will be able to generate code based on its understanding of that design information.
AI will become a tool for passing on experience
In the world of visual inspection, there are many things I’ve learned over the course of 10 years on the job.
For example,
- Lighting Techniques
- Concepts of Positioning
- Test Sequence
- Exception Handling
- How to Integrate with Equipment
Such experiences were, in the past, tacit knowledge possessed only by veteran engineers.
However, by organizing that experience as design assets and combining them with AI, we can pass that experience on to the next generation.
AI is not meant to replace humans, and
It is beginning to serve as a tool for reusing and further developing the technologies that humans have built up over time.
Future Competitiveness Lies in “Assetization”
As AI-assisted coding becomes more widespread, the differences in coding speed will likely narrow.
What will give us a competitive edge at that time is,
The quality of the design assets we possess
It is.
To organize my many years of experience and preserve it as knowledge that anyone can use.
We believe that this will serve as the foundation for new software development.
ovit, we are currently reviewing each of the visual inspection technologies we have developed over the years and restructuring them to suit the AI era.
Software development isn’t something you do once and then forget about. We believe that passing on the technical skills and know-how cultivated in the field over many years—along with AI—to the next generation is a crucial role of development moving forward.
