AI and Machine Learning Explained

In this context, we are primarily discussing machine learning, a subset of AI that analyzes extensive datasets to identify patterns and make predictions about future behavior. For instance, Amazon uses these algorithms to predict purchase preferences, while YouTube employs them to suggest videos.

Unbeknownst to many, machine learning models permeate various aspects of our lives beyond just e-commerce and streaming services. They play critical roles in fraud detection, security, and marketing, among other applications. However, these systems are limited by the data they are trained on. When faced with situations that deviate significantly from their training datasets, they struggle to make accurate predictions.

The Pandemic's Disruption to Algorithms

As COVID-19 began to influence daily life, machine learning algorithms encountered notable challenges. According to MIT Technology Review, one company experienced a breakdown in its predictive algorithm due to an unusual surge in bulk purchases. The recommendations for restocking items became misaligned with actual customer behavior.

Similarly, an increase in users caused discrepancies in the recommendation systems of a streaming service. The sudden influx of data led to less precise suggestions for users. In the case of a credit fraud detection firm, it wasn't the volume of purchases that raised concerns, but rather the nature of what was being bought. Consumers suddenly began purchasing gardening tools and power equipment, which the algorithms typically flagged as potential fraud.

Fortunately, the engineers at this company recognized the shift in consumer behavior and made necessary adjustments. Such modifications are vital for maintaining the functionality of machine learning algorithms during significant changes. Automated systems cannot simply be trained and then left alone; they require ongoing monitoring, adjustments, and sometimes retraining.

Adapting Algorithms for New Norms

For instance, Amazon adapted its algorithms to accommodate the shifts in consumer purchasing patterns. Typically, they prioritize sellers who utilize their warehouses. However, as online shopping surged during the pandemic, these warehouses became inundated with orders. By refining their algorithms, Amazon was able to distribute orders more evenly among sellers.

Another organization that employs machine learning for crafting marketing emails also adjusted its algorithms. Their system identifies optimal phrases to use, but given the current global context, they decided to avoid terms like "going viral."

As we navigate through and beyond the pandemic, the teams responsible for maintaining machine learning models will need to continue adapting to these evolving circumstances. This adaptability will ensure that algorithms accurately reflect the new realities we face.

Innovations in Machine Learning Due to the Pandemic

Moreover, the pandemic has spurred the development of new machine learning models aimed at addressing COVID-19-related challenges. With the ongoing uncertainty, scientists are designing algorithms that forecast the virus's trajectory.

For instance, a study released on May 19 by researchers at Mount Sinai Hospital integrated chest CT scans and patient data with machine learning models to diagnose COVID-19 cases. Their findings revealed that, in certain instances, their algorithm outperformed even experienced thoracic radiologists in identifying infected patients.

Other medical facilities, eager to understand which patients may deteriorate into critical conditions, are utilizing unvalidated algorithms to gain insights. Additionally, machine learning is being employed to model the spread of the virus and evaluate the effectiveness of mitigation strategies such as quarantines and social distancing measures.

Researchers at Rensselaer Polytechnic Institute have centered their analysis on smaller cities, recognizing that limited data availability complicates modeling efforts. Adjustments are essential to ensure accurate predictions in these instances.

As with algorithms that influence our everyday experiences, these models must be continually updated as we gather more information about the virus and adapt to new realities.

The Future of AI in Our Lives

The overarching goal of AI is to enhance our lives, and machine learning plays a significant role in achieving this across various sectors. However, with this interconnectedness comes the responsibility to evolve alongside these technologies. Failing to do so could result in them lagging behind our rapidly changing world.

The first video titled "From Lab to Life: What AI Tells Us About 'Long Covid'" explores how AI contributes to understanding the long-term effects of COVID-19.

The second video, "How AI and ML are Helping to Tackle COVID-19," discusses the role of AI and machine learning in addressing the challenges posed by the pandemic.