(Newswire.net — October 30, 2019) — There was once a long-standing debate in the computer science world. This was the dichotomy of “human-like AI” vs “simulated AI.” One side held that computers would one day be advanced enough to completely emulate human intelligence, since in theory once a sufficient neural network is strung together you have a replica of a human mind. The other side held that true human-like AI would always be impossible and that the best-case scenario would be to use proven computer strengths to brute-force some tasks in a limited space.
But there was one possibility that neither side of the debate could foresee: The evolution of deep-learning AI. Deep learning involves plunging an algorithm into an endless stream of data, with some minor guidance on what task it is to accomplish and no prior instruction on how to do it. Instead, the computer is given the freedom to make up its own rules in a trial-and-error process, as long as it gets the result we want. This was an impractical method even twenty years ago because memory technology did not have the capacity to store and sort such huge input and come up with its own answer.
The Deep Learning Revolution
Deep learning has been researched for practical tasks such as driverless cars and facial recognition, to great success. Then along came a joint venture of MIT researchers and Harvard Law experts in 2016, who founded the company Evisort. If deep learning could learn to drive a car or recognize a face, they reasoned, why not train one to read?
They didn’t propose training an AI to digest novels or anything broad like that, just legal documents. The language of legally binding contracts is a formal one, with predictable templates and repeated patterns. In many ways, “legalese” is very much like a programming language itself, being more engineering than prose. The research team discovered that an AI was very adept at picking up the patterns of legal contracts, and they launched Evisort that year as a cutting-edge document parsing application.
“Google For Contracts”
The Evisort system is a full office software suite that is designed for legal contract management. After a round of interviews with the likes of Forbes magazine and a round of investor funding backed by Silicon Valley’s biggest moguls, Evisort began attracting clients from all over the business world.
The point of legal contract management is to scan and mine data from a contract, identifying key data points. These data facts, such as dates, names, limits, terms, dollar amounts, and other pertinent details, can then be automatically stored in a database. Evisort’s system works in conjunction with other widely-used office tools, making for an agile platform that can digest 30-page contracts in seconds and integrate the data with standard office tasks.
The work that this deep-learning algorithm can do replaces hundreds of hours of human labor. The work of reading contracts up until now was tedious and prone to errors. In addition, modern technology has enabled the speed of business to increase, resulting in a paper deluge that can swamp even the largest law firm. The average attorney might handle dozens of contracts per week, to say nothing of the load put on in-house legal departments within existing corporations.
This data deluge is due to the fact that computers are very good at automatically generating data. A legal contract template can be set to be filled in with a list of terms, for instance. But the speed with which our modern automation can churn out data has far outmatched the number of eyeballs we can get to read that data. So any automated routine we can get to process contract data back into easily-navigated spreadsheets, and even automated beyond that within the system, is helping us keep up.
Ripe For Innovation
The company founders celebrated their success in a recent Harvard Law interview. Evisort founder Jerry Ting confides, “When we say, ‘Hey, in six seconds we can review a 30-page contract and pull out information for you,’ lawyers say, ‘Why did I spend 10 years of my life doing that?'” The legal industry is slow to change, moves at a conservative pace, and is sensitive to liability. Ting called the field ripe for innovation and has since guided the company to a customer base that encompasses Fortune 500 companies, Am Law 100 firms, independent entrepreneurs, and even sports teams.
Companies like Evisort have even restored a little faith in the computer science community. After years of researchers facing dead ends, the simple application of deep learning has become a “silver bullet,” making new benchmarks in tasks previously thought impossible for computers. If a human-like AI ever is in our future, it might very well be formed from a collage of deep-learning algorithms.