The past few years have witnessed a considerable interest, surge and technological advancement in Artificial Intelligence (AI), Machine Learning (ML) and Deep Learning (DL). This is primarily due to the birth and rapid embracement of cloud computing and open-source software, allowing unlimited storage of any type of data (structured, unstructured, semi-structured or binary) and providing infinite computational ability to process it at scale at an affordable cost with open source tools, software, and technologies. This is enabling rapid innovation.
While AI, ML and DL are very hot topics and are being utilized for data monetization across all industries ─ including the biopharmaceutical industry, where these tools are being explored to improve clinical development processes ─ the perception that these technologies are interchangeable has created some confusion.
In this article, we will review their definitions. As you’ll see, these technologies are not interchangeable but have distinct meanings with having discernible differences.
Artificial Intelligence (AI):
AI is human intelligence exhibited by machines. This is an area of computer science that emphasizes the creation of intelligent machines that work, act and react like humans do.
There are two types of AI:
- General AI: systems or devices that can handle any task. Although less common, this is where some of the most exciting advancements are happening today. This contains all characteristics of human intelligence including, but not limited to, planning, learning, reasoning, problem-solving, knowledge representation, perception, motion, and possible social intelligence and creativity.
- Narrow AI: Exhibits some form of human intelligence, but is usually focused on one specific task, such as image recognition.
Figure 1: Artificial Intelligence is a superset within which Machine Learning and Deep Learning belong.
Machine Learning (ML):
Machine Learning is an approach to achieve Artificial Intelligence. It is a subset of AI in the field of computer science that often uses statistical techniques to give computers the ability to “learn” (i.e., progressively improve performance on a specific task) with data, without being explicitly programmed.
There are various types of ML:
- Supervised learning includes systems that are exposed to large amounts of labeled data. Some systems may need to be exposed to millions of examples to master a task.
- Unsupervised learning includes systems that identify patterns in data, trying to spot similarities that split data into categories.
- Semi-supervised learning is a mix of supervised and unsupervised learning using small amounts of labeled data and large amounts of unlabeled data to train systems. The labeled data is used to partially train a machine-learning model, and then that partially trained model is used to label the unlabelled data, a process called pseudo-labeling.
- Reinforcement learning is concerned with how software agents should take actions in an environment to maximize some notion of cumulative reward. A computer system receives input continuously and is constantly improving.
Deep learning (DL):
Deep Learning is a technique for implementing Machine Learning. It is a subset of ML that has networks capable of learning in an unsupervised manner from data that is unstructured or unlabeled. DL operates on large volumes of unstructured data such as human speech, text, and images. A deep learning model is an artificial neural network comprised of multiple layers of mathematical computations on data, where results from one layer are fed as inputs into the next layer in order to classify the input data and/or make a prediction. Just like we use our brains to identify patterns and classify various types of information, DL algorithms are taught to accomplish the same tasks for machines. In contrast to ML, deep learning needs high-end machines and considerably big amounts of training data to deliver accurate results.
By understanding the similarities and differences among these emerging technological concepts, industry leaders – including those in the biopharmaceutical industry – are beginning their journey toward leveraging the benefits and applicability of AI and ML in improving efficiencies throughout their product lifecycles.
Please check back soon for Part 2 of this series where we’ll talk about machine learning techniques, best practices and algorithms to solve practical business problems.
Prakriteswar Santikary, PhD is the Vice President and Global Chief Data Officer at ERT.