In a recent development by IBM, they have come up with a high profile public demonstration of AI where they have used machine learning technology in the complex subject of organic chemistry. The AI chemist can predict chemical reactions in a way that could prove to be incredibly important for areas like drug discovery.

According to a recent paper published by the company, it is using highly detailed data set on thousands and millions of different reactions taken from various research paper published over the years. With this the researchers are embarking on the journey of making the AI use large quantities of knowledge that otherwise would have been a challenging task for humans to master.

Approaching the area of organic chemistry in an unusual way, IBM has used artificial intelligence in the same way as it does for translating languages. It is taking atoms as letters to predict outcomes of organic chemical reactions. By learning the syntax of reactions, it can predict the correct outcome almost 80% of the times. Though it is far from being perfect, it is still an incredible tool to cut down on the amount of time required to research millions of chemical reactions that have been previously documented.

Why is it needed?

Organic chemicals are highly complex and simulations of their behaviour can prove to be time consuming and may not even yield the desired result. Even the synthesis of pharmaceuticals and complex organic compounds can be a daunting task demanding 30-40 steps. This can be particularly challenging in the commercial sector, where they need to find shortcuts and skip a couple of steps to yield faster results.

The researchers are therefore using these algorithms for these chemical reactions to understand the basic structure of the language of organic chemistry.

How does it work?

In this new AI program, which is an artificial neural network, neurons are fed data to solve a problem such as translating a sentence. The paper further explains that this neural network then repeatedly adjusts the connections between neurons to see if the new pattern is better at solving the problems. Over the time it would figure out the patterns that work best at computing solutions by mimicking the process of learning in the human brain.

“It reasons and learns by analogy, which is very similar to what top pro organic chemists do in real life,” said Teo Laino, one of the researchers on the project from IBM Research. Still in the learning phase, researchers intend to feed the AI with factors like temperature, solvents, and pH so that it can learn better.

Instead of using AI in a way that can can replace human jobs, they are using it to augment the abilities of human being, as this effort can yield to spectacular development in the area of drug discovery. “We didn’t create this tool to replace organic chemists, but to help them,” said Laino.  

Though the tool is not publicly available right now, the company plans to make it available through a cloud service.