The frenzy around artificial intelligence (AI) has reached an all-time high following the success of ChatGPT. The market is flooded with all sorts of products claiming to solve problems with AI. However, effective AI products can’t be built simply by taking your favorite neural network and applying it haphazardly to a given problem. It takes domain expertise and an intimate knowledge of the training data before a credible and useful AI product can be built.
First, let us talk about our expertise on the subject. ULINK has been at the forefront of the drive testing industry for over 20 years. One huge pain point for our clients and partners was in the maintenance and effective utilization of HDDs. There was a need for an easy, reliable way to identify the drives at risk of failure. So around 2019, we set out to solve this pain point using machine learning, and more importantly, our knowledge of what each drive log and drive health attribute potentially meant for drive failures.
Every day, we collected data on several health attributes to understand the potential symptom patterns associated with drive failure. In two years, we were ready with a solution called ULINK DA Drive Analyzer that was compliant with California Consumer Privacy Act (CCPA) and EU General Data Protection Regulation (GDPR).
ULINK DA Drive Analyzer is powered by a cloud AI engine that collects and monitors data from drives to intelligently predict if one or more of your drives are near failure. Its user-friendly interface can be accessed via a web portal, desktop-based application, or NAS-based application, making it easy to monitor drive health status. Email notifications inform users when signs of deterioration appear, allowing them to back up data or replace their drives before catastrophic failure strikes.
So, what do we think makes an AI product credible?
Problem Solving Experience
To be able to create an AI product, one needs to have the relevant experience in solving a particular pain point and knowledge of the relevant processes involved in arriving at a solution to it.
Intimate Knowledge of the Training Data
The first thing that an AI model needs to be trained to solve a problem is access to relevant data. It is also important to note how the data was accessed and whether Privacy laws were being respected.
Data and how you use it are the driving forces behind all AI solutions. With the availability of data comes the potential to train AI solutions. At ULINK we have deep knowledge of such data, and prior experience tackling problems within the domain of storage drives.
QNAP and ULINK Release DA Drive Analyzer, AI-powered Drive Failure Prediction Tool for NAS
Photo Credit: monsitj