
Speaking virtually at the AI & Big Data Expo in Amsterdam yesterday, Vaibhav Verdan – the pharmaceutical company’s Global Advanced Analytics leader – noted that the speed at which vaccines have been brought to market was in a large part thanks to AI and ML algorithms.
“By March last year vaccines for human trials were already in progress – that was just three to five months after the first reported case in Wuhan.”
“AI and ML models helped researchers crunch vast amounts of data -across geographies and across structured and unstructured data, ” he explained.
According to Veran, this enabled researchers to increase their speed and accuracy, allowing scientists to apply deep learning ‘LinearFold’ algorithms -which are around 100 – 200 times faster than traditional algorithms – to make protein folding predictions that were key to fighting the virus.
“We know that there are probably tens of thousands of components attached to a protein. So the machine learning model was able to cut through this blizzard of data and give direction to the researchers and help them predict the most important components or subcomponents.”.
Verdan also praised advances in computer vision and image processing technologies that has enabled doctors to distinguish between cancerous cells and benign cells by looking at images. “Those algorithms are much faster and more accurate than a human and enable doctors to make a better diagnoses,” he said.
However, despite these “world saving” advances and the huge strides that have been made in enterprise AI over the past eight years, Verdan – who has also worked for utility and retail companies, warned that one of the challenges that all industries face now is the lack of quality data.
“Having access to good quality data and complete data which is refreshed is one of the biggest challenges we are facing,” he said.
Fellow panellist Surajit Basak, director of Information Technology at logistics firm UPS agreed on this point – and advised firms to “take a step back “ and make sure they had the basics in place before applying AI to their business intelligence roadmap.
He said: “For enterprise-wide organisations the first challenge they face is not AI but big data. How do they handle the data? What data is big data and what data is mud? They need to build a data warehouse to start this journey.
“The second part of the process is deciding where to store your data – especially in a hybrid model – do you put on cloud and what you leave on prem? And then there is the question of identity management and who you give access to your data. These are all steps you need to make before you even start on AI and predictions.”