IBM’s Watson was put in a particularly bright spotlight in 2011 when it faced off against the world’s two most successful game show contestants of all time on Jeopardy. When it came to the Final Jeopardy question, Watson comically gave the response “What is Toronto?????”
The category was U.S Cities.
Nonetheless, Watson won the match by more than double his opponents’ scores combined. Despite the occasional hiccup, Watson would go on to beat Jeopardy superstars Ken Jennings and Brad Rutter twice more. Since then, the world’s imagination has run amok with the possibilities and applications for machine learning. This has led the corporate world into a funding blitz.
In the last decade alone, venture capital firms have invested over $61 billion into AI companies. The tech giants have jumped into the race as well. Alphabet, Google’s parent company, has acquired at least 19 AI firms this millennium, putting it three ahead of rival company Apple.
Given the way enterprise systems are evolving, it’s not hard to see why businesses would focus so much on machine learning models. From social media platforms to entertainment services such as Netflix and Hulu, modern technology is built on “Systems of Intelligence.” This brainy paradigm is composed of recommendation engines, automation schema, and business insights centered on data analytics.
As businesses try to keep up in this evolving landscape, many analysts have tried to guess where AI will go in the future. Some big promises have been made. Namely, Elon Musk has committed to putting fully self-driving Teslas on public roads. The more time you spend online, the more outlandish the predictions you’ll discover. Most of these arise from a misunderstanding of AI, machine learning, and data science. In order to make accurate predictions about the future of data science, some myths need to be refuted.
Myth #1: Only a Data Scientist can contribute to machine learning solutions
When you look under the hood, machine learning models can seem quite daunting. Sure, there’s advanced math, statistics, and programming involved, but that doesn’t mean that there’s no role for someone outside of these vocations. In fact, a shortage of data scientists prompted the creation of what has been called the “Citizen Data Scientist.”
More and more technical implementation of data science technologies has become automated, and the technologies themselves incorporate much of the advanced mathematics required to properly execute machine learning. This certainly hasn’t diminished the roles of the data scientists and the engineers, but it allows business stakeholders to contribute as well. With this automation, also comes some danger. All models are built on assumptions and a detailed understanding of the underlying assumptions is critical for effective application of the models.
That’s precisely who the citizen data scientist is. They are business stakeholders with enough technical background to help the company make better decisions through data. Any data science team needs a subject matter expert to keep the project aligned with its goals, as well as someone with the capability to understand the insights gathered from the models and guide the organization in translating the model insights into effective business actions.
Data science isn’t just for technical majors anymore. Many of the top MBA programs are introducing masters in Data Science/Business Analytics, and multiple statistics and data science courses are required for many undergraduate business degrees. What used to require a large team of mathematicians and engineers can now be accomplished by smaller groups of more diverse specialists. This, combined with the automation brought by machine learning models, may have had a hand in the creation of the next misconception.
Myth #2: Implementing machine learning will take away human jobs
It is true that AI will lead to the obsoletion of many roles. Oxford Economics predicts that up to 20 million manufacturing jobs will be replaced by robots by 2030. However, AI should not be seen as the Grim Reaper of careers, but rather a transformative entity that creates new opportunities. The World Economic Forum, for instance, expects that automation will displace 75 million jobs but generate 133 million new ones worldwide by 2022.
The truth is that AI and machine learning are complementary to human skills. People are not good at quickly analyzing large volumes of data for patterns or inconsistencies. Computers are good at, well… quickly analyzing large volumes of data for patterns or inconsistencies! Humans are better with building relationships, making connections between seemingly unrelated things, and decomposing complex problems into simpler ones. While there will be some job displacement due to automation, there will also be considerable job creation and increased reliance on knowledge workers.
Myth #3: A business can’t implement machine learning solutions with low quality data
Data is king. There’s no denying this ubiquitous industry phrase. It is true that effective machine learning models need to be trained on high quality data because these models can only predict as well as the historical information we use to build them. However, we must start the journey of implementing machine learning in order to improve data quality and identify data gaps.
In many cases, building models can help identify gaps in data, so it’s important to start on that journey. Even with the most robust datasets, we are always identifying areas that can be improved upon. That said, if we at the very least get a moderately predictive model in place, higher quality data can be collected for future iterations of our models.
Every iteration should have both predictive goals and data improvement goals. When these goals are met, the initial model can be improved and further data gaps can be identified. Machine learning models are iterative by nature due to rapidly changing business processes, so it’s key to start on the machine learning journey and develop a clear AI strategy and roadmap.
Myth #4: Machine learning models are a black box that provides silver bullet insights
It may sound intuitive, but it’s important to understand that machine learning solutions won’t necessarily, or even usually, give clients major insights into their businesses that they don’t already know. The machine learning solution will never know your business as well as you do. A deeper understanding of the underlying science is important in realizing the value in data science and advanced analytics.
Earlier we discussed how automation has transformed data science teams. Advances in AI technology have also given rise to auto-machine learning and no-code analytics tools that reduce the barriers to entry for data science. These tools are enabling low or no code modeling and will become more sophisticated and more widely available over the next several years.
“Easier predictive models” doesn’t mean “better predictive models.” If a model is applied incorrectly, it can lead to inaccurate or incorrect interpretations. This is where Atrium helps businesses on their AI journeys. If we as a team have done our job right, we built the model by doing a deep dive into the business processes and aligning the model with business strategy. This takes a greater understanding of the business, the questions they want to answer, the data, and much more.
Increased understanding and expert support from Atrium with Elevate
The explosion of data science into modern boardroom vernacular has led to several misunderstandings regarding the field. As we move toward the future, it’s important that we build confidence today in what AI and advanced analytics can and can’t do. With this increased understanding, we will be able to design processes around data science and amplify employee productivity.
Through our Elevate service, we are providing customers with ongoing support and enhancements of their data science and analytics solutions with dedicated resources with the essential capabilities. Learn more about Elevate and find out how we can help you succeed with your future data science endeavors.