RAND researchers interviewed data scientists and engineers with experience in building artificial intelligence and machine learning (AI/ML) models in industry or academia to investigate why AI projects fail. They synthesized the experts' experiences to develop recommendations for smart implementation of AI. The lessons from earlier efforts to build and apply AI/ML will be helpful for others to avoid the same pitfalls.
Authors
- Division
- RAND National Security Research Division Acquisition and Technology Policy Program
- Pages
- 20
- Published in
- United States
- RAND Identifier
- RR-A2680-1
- RAND Type
- report
- Rights
- RAND Corporation
- Series
- Research Reports
- Source
- https://www.rand.org/pubs/research_reports/RRA2680-1.html
Table of Contents
- The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed 1
- Methods 4
- Industry Participants 4
- Academia Participants 5
- Findings from Industry Interviews 5
- Leadership-Driven Failures 5
- Many leaders are not prepared for the time and cost of acquiring cleaning and exploring their organizations data. 6
- Bottom-UpDriven Failures 7
- Data-Driven Failures 7
- Failures Due to Underinvestment in Infrastructure 8
- Failures Due to Immature Technology 8
- Two Special Cases Compute Power and Availability of Talent 9
- Understanding which problems are a good fit for AI and which are at or beyond the current state of the art can help organizations avoid costly and embarrassing failures. 9
- Agile Software Development and Artificial Intelligence 10
- Industry Interview Takeaways 10
- Results of Interviews with Representatives of Academia 11
- Activity Prestige 11
- Improper Data Structures 12
- Publication Incentives 12
- Other Findings 13
- Academic Interview Takeaways 13
- Industry Recommendations 14
- Ensure That Technical Staffs Understand Project Purpose and Domain Context 14
- Choose Enduring Problems 14
- Focus on the Problem not the Technology 14
- Before they begin any AI project leaders should be prepared to commit each product team to solving a specific problem for at least a year. 14
- Invest in Infrastructure 15
- Understand Artificial Intelligences Limitations 15
- Academic Recommendations 15
- Overcome Data Collection Barriers Through Partnerships with Government 15
- Expand Doctoral Programs in Data Science for Practitioners 15
- Industry Interview Template 16
- Academia Interview Template 17
- Notes 18
- References 19
