AI project failure rates top 80% — study cites poor problem recognition and a focus on latest tech trends among major problems

Everyone and their aunt seem to be hopping aboard the AI train in search of inflated profit margins and marketing hype — just look at AMD's recent Ryzen rebrand as a prime example of this AI hype. A recent study conducted by RAND has found that this AI-centric approach may not be all it's chalked up to be, with AI projects seemingly failing twice as often as regular software development projects.

During the study, RAND interviewed 65 industry experts with over five years of experience developing AI and machine learning tools for private entities and academia and distilled their responses into five main reasons for AI/ML project failures.

The number one failure, according to the study, was a leadership failure rather than a technical failure. Executives either failed to understand what the problem was that they were trying to solve with AI, failed to communicate the problem to development teams, or tried to apply AI to a problem it was ill-equipped to solve. Project leads were so focussed on using the latest and greatest AI advancements to solve their problems that they missed simpler, cheaper solutions that didn't use AI.

As one interviewee explained, his teams would sometimes be instructed to apply AI techniques to datasets with a handful of dominant characteristics or patterns that could have quickly been captured by a few simple if-then rules.

Resource availability was also a significant failure point, with leadership cited as being unwilling or unable to assign the necessary resources to process the necessary data and train AI systems adequately. This frequently results in a project underdelivering or delivering a product that was incomplete — a consequence of underestimating just how complex it is to create and train an AI system.

Similarly, many leaders had unrealistic expectations of AI as a result of recent hype and marketing claims, which becomes problematic when development teams are unable to deliver what was promised in the time frame expected of them.

For a more detailed look at the data, reasons for failure, and the researchers' recommendations, check out RAND's research report.

Ce qui précède est le contenu détaillé de. pour plus d'informations, suivez d'autres articles connexes sur le site Web de PHP en chinois!