VMBlog published an article composed by Civo Chief Innovation Officer Josh Mesout in which he discusses the trials and tribulations of the mission to make Machine Learning (ML) accessible to all.
Even though Artificial Intelligence (AI) is all the glitz and glamour in the tech world at the moment, with more and more new AI models being generated and rolled out like a mass-scaled production factory, ML has its own place in such conversations.
ML is a branch of AI that start-ups and enterprises have wanted to utilize for a long time, however, with the complications of cost, resources, and time necessary to create such infrastructure, it can be hard to understand the value when comparing the input to output. According to Gartner 85% of ML projects fail to deliver, and only 53% of projects make it from prototype to production.
Chief Innovation Officer Josh Mesout spoke about this more by saying:
As a result of these significant internal demands to run ML, more and more engineers are becoming dependent on open source to help resolve these issues. According to Anaconda's State of Data Science 2022 report, 65% of companies lack the investment in tooling to enable high-quality ML production, with 87% of organizations already leveraging open source software.
Smaller organizations face difficulties in reconfiguring their infrastructures to achieve ML goals as it is time-consuming and offers little rewards. This is particularly challenging during their growth stage when they focus on increasing margins and securing their future.
According to D Scully at Google Research, 60% of hours are spent on infrastructure engineering, 20% on data engineering, 15% on feature engineering, and only 5% on ML engineering.
To drive the accessibility of ML to the next level, a constructive ecosystem needs to be built and maintained. The users and resources are already there but need to be channeled in the correct manner. Investing in open source cloud ecosystems can remove barriers to the adoption of ML and make it more accessible.
Organizations are turning to open source projects as a solution to overcome these challenges. Open source delivers the most cost-effective way to run ML algorithms by providing readily available non-proprietary infrastructure, eliminating the need to invest time and resources in learning tools like AWS SageMaker before achieving ML Insights.
Open source technologies not only offer on-demand accessibility to ML, but also provide the same benefits as the most popular and high-quality tooling. Additionally, these technologies can be customized to fit specific use cases with a set foundation, reducing time consumption and complexity.
Kubeflow is a great example of the above. It is consistently contributed and maintained by some of the best experts in the business, allowing organizations to leverage knowledge they may not have internally due to size or a different business focus. This is particularly beneficial for companies that could strongly benefit from ML open source.
ML shouldn't be a closed shop, where its potential is only realized by those of scale. Through prioritization of developer's needs and open source tooling, ML can be made accessible to all.
Investing in open source cloud ecosystems is crucial to remove barriers and making ML more accessible. By creating an ecosystem that promotes growth and development for everyone, we can further the democratization of ML.
Developing more interoperable tooling will benefit developers and make it easier to advance their work in open source, ultimately benefiting the entire community. Improving GPU solutions, such as GPU Edge boxes, can also reduce barriers to adoption by enabling effective ML in multiple use-case placements, making them ideal for in-house workloads.
GPU instances, with their fast launch times and bandwidth pooling, provide a streamlined effort that enables the ability for organizations to have transparent pricing models. This is due to a reduced number of unknown costs that can surprise smaller companies, therefore, impacting their bottom line and budget, which could be used for better ways for development.
Fractional GPU instances, in particular, are a great gateway into the ML world for smaller, lower-scale businesses or hobbyists due to their similar benefits to GPU instances but in a smaller, more manageable way. This exposes those to the open source community or world of ML that would not be able to access it any other way, which is only a positive for ML as we look to diversify with different knowledge, backgrounds, and perspectives.
Check out VMblog to learn more about the article or Josh Mesout.