By 2017, Artificial Intelligence was moving quickly from research into more industries. I believed our education systems needed to respond, even though traditional institutions could not always change at the same speed as technology.
At the time, my company, RedLotus, was exploring deep learning in our development lab for advertising technology. I had also completed a mentorship series with young people in a struggling education system. Their fascination with what AI might make possible brought energy into the room. In 2017, smart cars, disease forecasting, and election modeling were among the possibilities researchers and companies were exploring.
A bit too much science fiction for you? Much of it was already being tested. Machine-learning systems used computation and data to estimate the probability of an outcome. In 2017, the emerging-technology list included self-driving cars, biometrics, chatbots, drones, 3D printing, and virtual reality.
As with anything exciting, AI invited hype. The better response was to understand what was possible in 2017, what was still developing, and how to prepare for the change.
What Is Machine Learning?
In simple terms, machine learning uses algorithms to identify patterns in data and improve performance on a task without every rule being explicitly programmed. As new data arrives, a model can be updated to account for patterns that changed.
How did I believe this would affect us? Basic spreadsheet analysis alone would not be enough for many machine-learning roles. The field demanded stronger foundations in linear algebra, probability, statistics, multivariable calculus, and computational optimization.
That is why I argued in 2017 for a stronger focus on science, technology, engineering, and mathematics in school curricula. Linear algebra, vector spaces, norms, probability, and optimization form part of the mathematical foundation used to understand many machine-learning methods.
If that is Machine Learning, then what is Deep Learning and Artificial Intelligence?
Good question. Let us dissect this over a time lapse.
Artificial intelligence, machine learning, and deep learning are related, but they are not interchangeable. The 1956 Dartmouth workshop is commonly treated as a founding event for AI as a field. By 2017, decades of research had turned that academic ambition into a major area of technology development.
The growth of large data sets and the use of graphics processing units helped accelerate parts of AI research. Machine learning, whose history predates the 1980s, can be viewed as one approach within AI. Algorithms learn patterns from training data and use them to classify, estimate, or predict. Methods include decision trees, clustering, Bayesian networks, and many others.
Deep learning grew out of earlier work on artificial neural networks and gained new momentum around 2010 as data, computing power, and training methods improved. Neural networks were partly inspired by ideas about how brains process information, but they are mathematical models, not replicas of the brain. Their layered representations can help machines perform tasks such as recognizing images, processing language, and making predictions.
In 2017, Silicon Valley companies were applying deep learning to shopping, ad targeting, driverless cars, and healthcare. It was also a research focus for RedLotus, the marketing-cloud company I was building at the time.
Test-ability
There were many test cases for deep learning. I knew the advertising-technology sector best. Large technology companies were exploring social, image, and natural-language applications. At RedLotus, we were testing deep learning in a controlled domain to study whether it could improve estimates of a consumer’s potential purchase.
The test examined signals such as past browsing behavior, transactions, ad viewability, clicks, social engagement, sharing habits, and other correlated events.
Machine-learning algorithms can model nonlinear relationships among variables in a data set. Those relationships can then support estimates about an outcome—for example, whether a user showed greater affinity for one vehicle model than another. The result was a probability, not certainty.
Machine learning can enhance predictive modeling. In 2017, many technology companies still used linear models because they were simpler to analyze and explain, even when more complex algorithms could capture relationships those models missed.
At the time, the industry was struggling to find enough people who could interpret these systems carefully. Mathematical intuition mattered, but so did testing, judgment, and an honest understanding of a model’s limits.
Prepare Yourself to Lead this Change
Personally, I do not believe education has to come only from a four-year college or university degree, although I strongly encourage people to attend college when they can. Institutional learning is not the only way to acquire knowledge. In 2017, universities and online-learning platforms were already making technical courses available at low or no cost.
To get your feet wet, you do not need to master all the mathematics before you begin. Start with an introductory course and build your foundation. If your passion grows, keep investing your time. Learn the mathematics as you advance. Spend time with deep learning and artificial intelligence so you can understand the tools before trying to build intelligent products.
Educational resources I used in 2017:
Coursera
Udacity
Khan Academy
Those were three starting points I used. Course availability and pricing change, but the lesson does not: explore structured resources, teach yourself the basics, and keep going.
When I was growing up, resources like these were not available to me. Take advantage of what you can access. Study on weekends, in the evening, or whenever your time permits. If you have the drive, passion, and desire to lead the next wave of digital innovation, invest additional time in your own growth.
If we waited only for government institutions to change the curriculum, I feared students and entire regions would be left behind. India, China, and the rest of Asia had the talent and ambition to help lead the next entrepreneurial and technical chapter.
The Future Awaits
In 2017, I saw artificial intelligence as a defining wave of the future. My recommendation was direct: prepare for it as early as possible. I also urged education leaders to strengthen the curriculum so young minds could understand and help build what came next.
The tools will change. The responsibility will not. Learn the foundation. Test what is possible. Prepare yourself to lead.
