Machine Learning in 2020: What to Expect from a Maturing Technology
It now seems clear that machine learning will play an important role in our daily lives. Already, machine learning models are at the heart of facial recognition, voice technologies, text scanning, product recommendations, and more. In the future, this impact will only grow as machine learning applications emerge across industries.
The challenge, of course, is that machine learning is still an early stage, an emerging technology. Many companies want to take advantage of AI and its potential benefits. However, the barrier to entry can be difficult for building data pipelines, effective models, and production architecture for these solutions.
In 2020, the biggest trends in machine learning revolve around making machine learning more accessible. This article will explore the various ways machine learning’s impact is growing and becoming more widely available for companies across industries and around the world.
Costs Decreasing, But Not Fast Enough
One of the biggest challenges facing new machine learning initiatives is cost.
Companies trying to implement machine learning often find that the budget for developing a solution is quite high. A large portion of the cost is attributable to the false starts and missed attempts that happen early on in the project’s life cycle.
Hiring for data and AI expertise is not cheap. Additionally, these models tend to be hardware hungry, meaning infrastructure costs can be high. Most machine learning initiatives won’t produce desirable results for the first few months of development at least, so the budget to create a new project can be a major challenge for companies. Especially smaller companies with less capital.
Thankfully, the trend is toward cheaper costs in machine learning. More technology gets open sourced every year, leading to faster time to market for end applications. In addition, computing resources are getting cheaper as Google, Microsoft, Amazon, and other companies compete to provide cheaper cloud infrastructure.
Better Models, Freely Available
With every year that passes, machine learning models get better. Critically, the machine learning ecosystem has a great culture of sharing and collaboration. As a result, most of the big innovations happening in machine learning are quickly released as open source code for anyone to use.
As we enter the middle stages of machine learning’s maturity, more good models are getting open sourced every year in all kinds of domains. No matter what machine learning application you want to build, chances are there exists an open source model or framework that will make your job easier.
Of course, PyTorch and TensorFlow are the most well known open source resources in the machine learning community. Every year, the list of great open source machine learning projects grows.
More Developers & Data Experts
One side effect of growing interest in machine learning is the growth of data and AI expertise in the marketplace. Machine learning has begun to hit a turning point where data science and model building skills are available and easy to learn for many developers.
Don’t underestimate the importance of having a large community of developers with training and experience in the domain. If machine learning is to become widely available to companies of various sizes, then a deep talent pool is an important contributing factor.
In 2020, the talent pool will continue to deepen as more developers realize that future job prospects will rely heavily on data and AI expertise.
Challenges for Deployment & Scaling
New machine learning applications face difficulties when it comes time to deploy to production environments. While it’s fairly easy to train a model on static data with open source tools, it’s still non-trivial to create and operate a data pipeline and workflow.
Add in the complexities of streaming data and multiple models processing the same messages the same way. Ultimately, machine learning applications can become quite complex to deploy and maintain.
Furthermore, scaling machine learning models to thousands or millions of users successfully comes with its own massive challenges. While companies like Amazon, Netflix, and Google have the expertise and capital to dedicate to solving these deployment challenges, smaller companies are having difficulty operationalizing their machine learning models.
2020 will see a deepening of the technology infrastructure for deploying and scaling machine learning models. However, don’t expect all the kinks to be ironed out by the end of next year. The truth is that machine learning models and data pipelines are inherently complex. They will always require expertise to implement and productionalize.
Better Cloud ML Infrastructure for All
That said, there are improvements that can be made to infrastructure to make machine learning models even easier to deploy. As the market grows and matures, expect cloud providers like Amazon, Google, and Microsoft to prioritize creating infrastructure solutions that are easy to deploy and scale.
Ultimately, much of the work of machine learning will happen on the cloud, instead of on-premise. The reality is that machine learning models can run much more effectively on specialized hardware. With their ability to provision, maintain, optimize, and upgrade hardware reliably, cloud providers will be the obvious choice for anyone building a serious machine learning application.
This is good in the long term because competition will drive down costs, allowing access to machine learning for even the smallest companies. Ultimately, even individuals can create, train, and provision resources for their own machine learning models to solve the world’s problems. That’s exciting for the democratization of AI.
2020 Is a Big Year for Machine Learning
At its current rate of growth, every year is a big year for machine learning year-over-year. 2020 feels significant because it marks the early stages of a maturing technology. It’s now clear the types of problems machine learning is well-suited to solve. Now, we can focus on making those solutions easier to create, deploy, and maintain for companies big and small across industries and around the world.
Founded in 1991, Intertech delivers software development consulting and IT training to Fortune 500, Government and Leading Technology institutions. Learn more about us. Whether you are a developer interested in working for a company that invests in its employees or a company looking to partner with a team of technology leaders who provide solutions, mentor staff and add true business value, we’d like to meet you.