Artificial Intelligence is here, but what is it exactly? How is it used and what is its potential? What is AI’s role in life science, and will it live up to its promise to help us reach our goals faster and more efficiently than ever? While the technology is exciting and promising for the future, it has yet to fully gain our trust, and many questions and doubts still linger about AI and its benefits. BioStock has decided to dedicate an article series to try to answer some of these questions, with respect to AI in the life science sector. With this first article we try to answer the question: what is AI?

“Mark my words, AI is far more dangerous than nukes.” Those are the scathing words of tech entrepreneur Elon Musk who worries about artificial intelligence (AI) taking over and becoming too “smart” for humans to handle in the future. Yet, for now, the rise of AI seems unstoppable as more and more companies, including Musk’s own electric car company Tesla, invest in the new technology.

One of the great appeals of a Tesla is the fact that it can drive itself, without the need of a driver steering the wheel or pushing the pedals. This seemed impossible just a few years ago, but now it’s quickly becoming a reality in many cars. But how does the car know when to turn? And when to brake? And how does it avoid other cars? The short answer: AI.

But what exactly is AI? With this article, we strive to breakdown the basics of AI and its subcategories.

A brief history of AI

Artificial intelligence is the computer simulation of the processes typical of human intelligence. These processes include learning and reasoning. Yes, machines can learn and think, apparently.

Arthur Samuel demos his checkers program to the public in 1956 (source: IBM)

As an academic discipline, AI was founded in 1956, and, in 1959, MIT computer science engineer Arthur Samuel coined the term machine learning. Already at that point, the goal was to get computers to behave as humans and thus become “intelligent.”

Samuel implemented AI to create computer games such as checkers. Just think of your checkers or card game on your computer – the computer simulates an opponent to play against. Nowadays, AI has advance so far that the seemingly simple checkers games on your iPad or cell phone learn from our moves and make the gaming experience ever-more realistic – as if we were playing against a checkers champion able to predict our every move. Scary, yes, but engaging and fun.

Machine Learning

AI is largely based on complex mathematical algorithms that are coded into a computer software. The algorithms read and compute huge amounts of data in order to give a computer the ability to predict a specific result. The more data the computer has to work with, the more “intelligent” it becomes.

The way in which the data is collected and computed determines how AI is subcategorized. Some forms of AI are smart enough to learn how to interpret data without input from a person telling the software what to compute. This form of AI is what we call machine learning today. In short, this type of AI not only mimics human behavior, it also mimics how humans take in new information.

Think of a weather forecast – meteorologists try to give us an idea of what weather to expect the next day or coming days and they are able to do so thanks to advanced software algorithms that use loads of weather pattern data from the past to predict how the weather will change in the future.

A similar kind of prediction software is used in diagnostic tools – with enough data on hand about a certain disease, let’s say cancer, so that medical professionals could implement a software to accurately predict if a patient has cancer from a blood sample, instead of using invasive measures such as surgery to collect a biopsy.

Follow us through this article series to learn more about such AI methods being currently used by the biotech industry right here in Sweden.

Machine Learning vs Deep Learning

Machine learning works by collecting and organizing data in several layers. As the complexity of the layering increases, a more powerful type of machine learning is required: deep learningThe examples mentioned above need powerful deep learning software to discern between the layers of data to make their respective predictions. Deep learning is thus a subset of machine learning that is able to separate types of data into different categories and make predictions based on those differences.

Deep learning is so complex and involves so many layers of data that its digital structure mimics that of a brain, where thousands, if not millions of connections are made between neurons each second. Instead of actual neurons, deep learning is based on groups of algorithms called artificial neural networks (ANN’s) that communicate with each other as new data comes in. So, deep learning software takes data connections between all artificial neurons (algorithms) and adjusts them according to a specific digital pattern as the learning progresses.

AI in life science

The advent of AI has quickly emerged as a big player in the life science industries, making huge impacts already. The emergence, fueled largely by the increasing availability of large amounts of data and cheaper, more powerful, computers, is generating great strides in innovation within all fields of life science, accompanied by major business deals between biotech startups and Big Pharma.

In our next instalment of this AI series, we will take a closer look at the inevitable rise of AI in life science, especially within the context of drug discovery, drug development and diagnostics. We will also go into some of the most important deals within the field.

Find parts II and III of this article series here:

The Applications of AI in life science

Overview of Swedish life science companies using AI

Prenumerera på BioStocks nyhetsbrev