Jump to content

Draft:ISO/IEC 5388 AI system life cycle processes

From Wikipedia, the free encyclopedia
  • Comment: None of these appear to be secondary sources. Stuartyeates (talk) 05:21, 29 April 2024 (UTC)

The ISO/IEC 5338 standard on AI system life cycle processes.[1] is an internationally agreed description of processes and concepts for the life cycle of AI systems. It builds on ISO/IEC/IEEE 15288 (system lifecycle) and ISO/IEC/IEEE 12207 (software lifecycle), by discussing what is different for AI systems[2]. These amendments are in part based on ISO/IEC 22989 and ISO/IEC 23053.

Organizations can use the 5338 standard as guidance for developing AI systems to significantly improve quality, cost and performance. It help developers and users of AI systems to gain clarity on how to responsibly design, develop and adopt AI systems, including elements such as risk management, stakeholder needs, maintenance, testing, validation, and disposal[3]

AI system development consists of many familiar processes plus processes that are AI-specific. For example, in machine learning part of the system behaviour is defined by training on data, instead of writing instructions in source code[4]. Furthermore, data may change, which requires continuous validation of the behavior with new data, and potentially retraining.

References

[edit]
  1. ^ "ISO/IEC 5338 page". ISO/IEC.
  2. ^ "NEN article discussinge ISO/IEC 5338". NEN.
  3. ^ "Get to know the global standard on AI systems". Software Improvement Group. 15 December 2023.
  4. ^ "Improving processes in developing AI systems". IEC.