The presentation focused on approaches to validation and how the FDA treats artificial intelligence technologies. The ultimate goal of AI is to get the computer to think and act like humans with the focus on thinking and acting rationally in order to be able to perform some of the functions we currently rely on humans to do. Currently, popular AI include Alexa, Siri and Sophia, all of which can do the simple task of looking for and retrieving information and sometime taking action such as playing a song. Yet, AI could do so much more.
Dr. Browner worked with neural networks and optimization in aerospace before founding Prelude Dynamics. She is currently developing an AI risk-based monitoring (RBM) tool to aid clinical study monitors in managing and monitoring the data. VISION™ Clinical Trial Optimization Platform already contains an innovative technology in its dictionary coding function. The function instantaneously matches the best options as the user types, allowing them to quickly select from a dictionary like WHO Drug, that contains over 300,000 entries. This IntelliMatch functionality is also used in search, reports, and exports as well as in the combo fields. Prelude is also planning to augment this technology with a learning algorithm to help offer the most prevalent options.
Dr. Browner said, “AI is going to change the landscape of clinical research. At Prelude we are strategically implementing our AI strategy and helping lead the industry into its implementation.”
Applications in artificial intelligence use methodologies to present knowledge, plan, process natural language, perceive things like speech, facial and object recognition, move (motion, as in robotics), and have social intelligence. One of the challenges, as an example, is that communication is known to be 80% or more non-verbal. So how do you teach a computer to understand that? The current strategy is machine learning.
Machine learning is the process of teaching a computer to learn by providing it with information and learning experiences. Initially, programmers teach the computer to learn through supervised learning. The end goal is for the computer to learn unsupervised. In order to accomplish this, technologies such as Baysian Networks, Neural Networks and Evolutionary Algorithms can be applied to assist decision making.