The survey found that 34% of those surveyed personally spend 0-10 hours configuring or setting up ML each month, with a further 24% spending 11-20 hours. Before developers can generate ML insights, they need to configure different aspects of complex infrastructure, such as machine resource management, monitoring, and feature extraction. This process can be very time and resource intensive, for only a small amount of ML insights. As a result, nearly half of those surveyed (48%) believed that ML projects are too time-consuming.
However, the survey also found that with open source tools, developers can be far more flexible and rapidly deploy ML and gain insights. 73.2% found that open source reduced the time from implementation to insight, with exact time savings for respondents including
46% saved 0-10 hours.
25% saved 11- 20 hours.
12% saved 21-30 hours.
12% saved 30+ hours.
The time pressures associated with ML deployment can be a significant factor in projects failing. 53% of ML developers abandon between 1-25% of projects, with an additional 24% reported leaving between 26-50% of projects and 8% between 51-75% of projects. Only 11% of developers said they have never abandoned an ML project.
Josh Mesout, Chief Innovation Officer of Civo, commented, “As machine learning is becoming more common place as a problem solving tool, we have noticed many developers who are being tasked with deploying ML are not ML experts. Instead, they’re domain experts in need of the insights and support of ML projects and tooling. However, the research shows that many are struggling to use this technology to its full potential, getting stuck in the surrounding infrastructure rather than reaping ML’s rewards.
“More needs to be done to highlight the benefits of open source tooling, which can significantly cut down on wasted time. There are a range of emerging services available that can help to offset the widely found pain points of ML. With access to open source, developers can tap into the ready-made resources created by ML experts and spend their time generating the insights they need rather than configuring the infrastructure to get there.”
