The calculations required to find new drugs are unimaginable to the human mind: it is estimated that there could be more potential molecules that should be evaluated than atoms in the solar system . Such a vast search would take a long time with current computer systems, but this is changing thanks to a new generation of hardware designed at the Argonne National Laboratory in Chicago.
Basically, the greatest advance has been made by deep learning algorithms (a class of algorithms devised for machine learning that have more layers). These algorithms are great for quickly finding patterns in large amounts of data. Now, along with these software enhancements, new hardware designs will multiply the possibilities exponentially.
Designs like the ones Argonne has announced that it is already testing on its new computer from the startup Cerebras and that promise to speed up algorithm training by several orders of magnitude. This computer is part of a new generation of specialized artificial intelligence hardware.
Beyond the GPU
Currently, the most common chips used in deep learning are known as graphics processing units (GPUs). These chips have traditionally been used in video games and computer-generated graphics production, so the biggest stumbling block with GPUs is that they are general-purpose: while they have fueled the artificial intelligence revolution of this decade, their designs are not optimized for such a task .
This inefficiency in hardware design limits the speed at which the chips can execute deep learning algorithms and requires enormous amounts of energy in the process . Also, many of the specialized chips are optimized for commercial deep learning applications, such as computer vision and natural language processing, but may not perform as well when handling the data types of scientific research.
By comparison, new chip architecture designs, such as the one introduced by Argonne, have the potential to train deep learning algorithms up to 1,000 times faster than GPUs , and using much less power in the process.
At the moment, thanks to the size of the Cerebras chip, which is larger than an iPad and has millions of transistors to perform large amounts of calculations, it is not necessary to connect several smaller processors, which would slow down the training of the model. In the tests , it has already been possible to reduce algorithm training time from weeks to hours . Most interestingly, Argonne has been testing the system in research to develop new anticancer drugs.
The goal is to design a deep learning model that can predict how a tumor would respond to a drug or a combination of drugs. This would involve developing an AI model to predict the properties of millions of molecular combinations . It’s like looking for a needle in a haystack, but thanks to this new architecture that improves not only the software but also the hardware , now everything can be done much faster.
Cerebras is not the only company that has since taken advantage of these new architectures: there are also startups such as Graphcore, SambaNova and Groq on the pitch. All this, in a very short time, could offer us insights in the field of medicine that otherwise would have taken decades to produce.
