Speed Reader: Evolution AI Accelerates Data Processing with AI

by Jack Watts

The prospect of processing monotonous mountains of paperwork is not what dreams are made of. Yet that’s exactly a task shouldered by many workers in the legal, healthcare and finance industries, among others.

Evolution AI, a U.K. startup and a member of our Inception program, has developed an AI platform that turns these mountains of paperwork into molehills, extracting and understanding information much faster than humans.

According to reading speed tests, the average adult can read 300 words per minute. With roughly 400 words fitting on a page of office paper, this means it takes more than 2 hours to read a 100-page document.

Evolution AI’s platform, which runs on NVIDIA GPUs, can read — and analyze — the same document in 2.4 seconds.

This frees up valuable manpower from mundane, manual data processing chores to concentrate on more complicated tasks.

And that’s not all this technology can do.

Recently, Evolution AI has worked with The Economist magazine to debunk fake news by analyzing thousands of tweets. It also improved how an insurance company assesses risk by autonomously reading hourly updates of U.K. court lists.

The company even joined forces with University College London and the U.K. National Health Service to research the effect of inflammation in mental illnesses such as bipolar disorder, schizophrenia and depression.

In previous studies of this kind, researchers have only been able to evaluate a few hundred psychiatric reports. Evolution AI’s GPU-accelerated AI platform was able to quickly pull key blood test data from 200,000 anonymized records.

The researchers will use this data to map how inflammation markers change at times of relapse and recovery, with the hope of proving the common view that inflammation plays a role in mental illnesses.

One Size Does Not Fit All

Unlike traditional natural language processing platforms, Evolution AI’s can tackle both structured and unstructured text. This is thanks to deep learning. Its GPU-accelerated platform doesn’t follow an explicit set of rules, but learns to understand the meaning of texts through observation.

“We did not want to create a one-size-fits-all platform,” explains Martin Goodson, chief scientist at Evolution AI. “We have therefore developed a platform which can learn for itself, by observing the work of human operators. This enables us to make a difference across industries and to be completely flexible.”

AI Podcast: How AI Can Improve the Diagnosis and Treatment of Diseases

We’re not going to live forever, but data-driven medicine promises to pick out subtle trends — patterns that may only be apparent when you can compare the health care histories of tens of thousands of patients — that can drive much more effective treatments.

Find out more in the AI Podcast featuring Mark Michalski, from the Center for Clinical Data Science at Massachusetts General Hospital, who discusses how AI technologies can help spot patterns that can improve the detection, diagnosis and treatment of diseases.