China’s state press agency, Xinhua, now features two artificial intelligence (AI) anchormen—one who speaks Mandarin and the other who speaks English. According to Xinhua, the anchors “can read texts as naturally as a professional news anchor.”
The AI anchormen were modeled after real-life reporter Zhang Zhao and were digitally created from footage of the real anchor alongside the script he was reading. Those assets were used as input, and then machine learning algorithms created additional voice and movements that were missing. In other words, the computer learned from Zhang Zhao and then filled in the blanks.
Machine learning is a noteworthy subset of artificial intelligence (AI). Simply put, machine learning enables a computer to learn from itself.
Back in 1959, MIT engineer Arthur Samuel coined the term and defined machine learning as a “field of study that gives computers the ability to learn without being explicitly programmed.”
Machine learning drives processes all around us. If you’ve ever seen a recommendation pop up while tagging a friend on social media, you’ve experienced machine learning. If spam is automatically filtered in your e-mail application, you’ve experienced machine learning.
Essentially, machine learning algorithms sort through large amounts of data and enable technology to make predictions based on patterns found in that data. Machine learning drives many of the adaptive learning functionalities that have become standard in learning management system (LMS) technologies. These functionalities enable us to see where, when, and how our learners are engaging with content, and guide the curation and recommendation of associated learning content and social media feeds.
EXTENDING OUR BRAINS
Using our human brains, we most likely can predict how a learner might score on an assessment based on past behavior. We can review enrollment and retention figures and predict how many of our learners will complete a program. We can use performance improvement metrics and compare “before” and “after” scenarios to assess knowledge transfer. With the advent of machine learning, we can do it all much faster and with greater accuracy.
Futurist and inventor Ray Kurzweil predicts that by the year 2029, computers will have human-level intelligence. As computers continue to improve and learn from themselves, Kurzweil posits humans will augment our abilities through the use of AI, forming a new type of hybrid processor. We see it almost happening today as we augment our abilities through smartphones, smartwatches, and computers, which Kurzweil refers to as “brain extenders.”
Even as we partner with our “brain extenders,” one thing remains constant: garbage in, garbage out. In other words, quality input is critical when developing and implementing machine learning models. And, in most cases, that input still relies on human intervention.
KEEPING A HUMAN IN THE LOOP
Human-in-the-loop (https://tdwi.org/articles/2018/07/09/adv-all-humans-in-loop-for-machine-learning.aspx) is a branch of AI that marries the best of human and machine intelligence to bring about better machine learning models. According to the team at Figure Eight (https://www.figure-eight.com/), “In a traditional human-in-the-loop approach, people are involved in a virtuous circle where they train, tune, and test a particular algorithm.”
With human-in-the loop, we (the humans) become responsible for tasks such as labeling data, fine-tuning models, and validating results. In this machine learning process, humans are valued as the stewards of data, expert decision-makers, and quality assurance team.
While “in the loop,” these practices can guide our outcomes:
- Explicitly define the questions you are looking to answer. Do you need to know which specific asset your learners are using? Or do you need to know which types of learning assets are being used the most? The machine will help you find the answers, but you need to establish the questions in advance.
- Find the stories in the data. Any time you have the opportunity to interact with data or results from machine learning functionalities, look for a story. Data may show your learners are asking a chatbot the same question at the same time every day. What is happening at that time? Ask your learners and find the narrative.
- Provide human perspectives alongside resulting data. When patterns emerge, add your perspective. Your knowledge is still a vital part of the machine learning process, and your perspective is a critical component of building better algorithms. You may know better than the technology can predict what is happening in your learning spaces.
- Tap into the data scientists. Chances are there is a data scientist lurking somewhere in your organization. If not, consider bringing one in to your Training and Development team. They can help you explore and interact with machine learning possibilities within your learning solutions.
As AI advances, futurists predict technology will, indeed, replace humans in design, development, and delivery of learning solutions. We can shape this future by having a better understanding of our own roles—as humans—in the machine learning process.
KEY INFLUENCERS TO FOLLOW
@Emerj: Artificial intelligence original market research and media. Professionals follow it to discover where AI is making a difference in their industry.
@FigureEightlnc: Figure Eight is the essential #HumanintheLoop #MachineLearning platform to create customized large-scale highquality #TrainingData.
@melissa_lamsonl: Author, speaker, leadership coach, and workshop facilitator, passionate about helping individuals, teams, and organizations thrive globally.
Phylise Banner is a learning experience design consultant with more than 25 years of vision, action, and leadership experience in transformational learning and development approaches. A pioneer in online learning, she is an Adobe Education Leader, Certified Learning Environment Architect, STC Fellow, performance storyteller, avid angler, and private pilot.