AI Models Have an Expiry Date — Continual Learning May Be an Answer
Why, in a world where the only constant is change, we need to take a Continual Learning approach to AI models.
Imagine you have a small robot that is designed to walk around your garden and water your plants. Initially, you spend a few weeks collecting data to train and test the robot, investing considerable time and resources. The robot learns to navigate the parts of the garden where there is grass and bare soil.
However, as the weeks go by, flowers begin to bloom and the appearance of the garden changes significantly. The robot, trained on data from a different season, now fails to recognize its surroundings accurately and struggles to complete its tasks. To fix this problem, you need to add new examples of the blooming garden to the model.
Your first thought is to incorporate new data examples into the training and retrain the model from scratch. But this approach is expensive and you do not want to do this every time the environment changes. In addition, you have just realized that you do not have all the necessary historical training data.
Next, you consider just fine-tuning the model with new samples. But this approach is risky because the model may lose some of its previously learned capabilities, leading to catastrophic forgetting (a situation where the model loses previously acquired knowledge and skills when it learns new information).
So…. is there an alternative? Yes, Continual Learning (CL)!
Of course, a robot watering plants in a garden is only an illustrative example of the problem. Later in this article you will see more realistic applications.
How To Ensure New Analytics Dashboards Get Adopted
When undertaking any kind of initiative, the importance of “starting with the end in mind” is a well-regarded principle famously attributed to author Stephen R. Covey. The end goal of investments into most new data and analytics (D&A) solutions is not to generate data-driven insights — it is to have users generate these insights for themselves. Optimizing for adoption is a crucial yet often overlooked success factor here.
In the Information Age, data is often referred to as the new oil. Enterprises are investing more than ever in solutions to expedite and enhance the process of extracting actionable insights from their data.
In our daily roles as consultants at Lingaro, however, we often see that not enough of those significant solution development resources are going into streamlining workflows. We frequently encounter situations where users are not taking full (or any) advantage of the latest and greatest D&A tools their employers have put significant time and money into developing.
What is behind this behavior? What can be done to avoid these types of scenarios? What are the challenges when it comes to using intricate analytical solutions? Is there a lack of proficiency in using these tools efficiently, or is there simply no necessity for such complex solutions? Could it be that the development process progressed too rapidly and did not factor in proper user experience (UX) design? This article aims to address these questions and provide insights into the pertinent issues.
Recommended by LinkedIn
The top reasons for poor adoption
There are many variables involved in designing and implementing new D&A solutions. One variable is obviously the scale of the investment. Another is the overall project approach taken, which might involve building applications in-house from scratch or hiring external vendors to build complex solutions for thousands of people around the world. And yet another is employee engagement when it comes to learning new skills.
Regardless of these variables, if a new D&A solution is poorly adopted then — in our experience — it almost always is because there was too much focus on delivering the product and not enough on embracing the design process.
De-emphasizing the design aspects of the project is like expecting someone who has never ridden a two-wheel bike before to be able to do so immediately once he has one. Disappointment is practically guaranteed.
Lingaro Leads Major Contenders in "Value Delivered" in Retail and CPG Data, Analytics, and AI Services PEAK Matrix® Assessment
In recognition of the company’s expanding capabilities to help enterprises in the retail and consumer packaged goods (CPG) industry succeed in all aspects of data, Lingaro has been named a Major Contender in Everest Group’s 2024 Retail and CPG Data, Analytics, and AI Services PEAK Matrix® Assessment. Lingaro was positioned highest among all Major Contenders in delivering value to its customers. This report evaluates data, analytics, and AI (DAAI) service providers based on various factors, including value delivered and the adoption rate of solutions among retail and CPG leaders.
Companies in the PEAK Matrix® Assessment are evaluated across two key dimensions: Market Impact and Vision & Capability. Market Impact includes subcategories like market adoption, portfolio mix, and value delivered. Among all the Major Contenders featured in the report, Lingaro was named the highest in “value delivered”. The company also received high marks in market adoption, overall market impact, and vision and strategy. Lingaro's strong performance in these areas underscores its commitment to delivering exceptional value to clients in the retail and CPG industries. Major Contenders are distinguished for having growing portfolios of DAAI transformation services, expanding capabilities to deliver these services, and proven track records of investments in proprietary tools and service accelerators focused on meeting the needs of RCPG enterprises. Everest Group highlighted Lingaro’s comprehensive set of technology partnerships, advanced technical expertise and talent management capabilities, and solid reputation among enterprises in the food and beverage as well as household and personal care CPG industries.
Click here to see the Everest Group Retail and CPG Data, Analytics, and AI Services PEAK Matrix®.
Got insights to share? Let us know in the comments.