Hyperion Research

Understanding AI

Written by Julie Potts | May 7, 2018 2:00:00 PM

For years, we have been hearing about the impact that Artificial Intelligence (AI) technology could have on the legal profession. The hype around how AI will “change the world” includes everything from promises of an AI-fueled utopia to the impending Skynet apocalypse. To say that expectations about AI are high would be an understatement and while much of the talk is around the possibilities, the reality is much less dramatic, is not always that useful or amazing, and can be extremely frustrating.

One should note that there is a clear delineation between hype around commercially available systems that use AI (which far exceed reality) and that surrounding academic research related to AI, where we've seen significant breakthroughs over the past couple of years in the accuracy of AI tools in making predictions and recognitions.

Despite all the hype, there are those that believe that AI has the potential to provide huge productivity gains and cost savings to businesses, freeing employees from those mundane tasks that a computer can handle and enabling them to focus more on those that add value.

Here, we aim to provide clarification and understanding around what AI is, how it works, and the technology today.

DEFINING AI

At the most basic level, artificial intelligence describes machines that use decision-making or computing processes that mimic human cognition.

Cognitive computing means teaching computers to learn, reason, communicate, and make decisions. Generally, this implies that a machine understands information about its environment and can then use that information to optimize actions that help achieve a specific goal. These cognitive tools are trained to complete tasks traditionally done by people with a focus on looking for patterns in data, testing the data, and finding results.

With the ever-increasing amount of data that organizations are gathering and storing, the ability of any human to review and comprehend it all without assistance is impossible. But human involvement is still very much an integral part of the technology as the data has to be given to the computer.

 

MACHINE LEARNING

Today, artificial Intelligence in business really translates to machine learning (a sub-set of AI), which can be a powerful tool for making highly accurate and actionable predictions. And, while much of the potential of AI is yet to be realized, machine learning is very much real and already here.

Machine learning is driving changes on three levels:

  • Tasks and occupations
  • Business processes
  • Business models

And while they don’t entirely replace humans, they complement human activities, making that work even more valuable.

With machine learning the computer is given data and begins to make decisions with minimal programming to predict an outcome. Algorithms (sets of instructions for solving particular problems) are utilized to allow the computer to determine the rules itself. Machine learning can recognize text, voice, and images. It is used daily when emails or messages are spell-checked; Siri and Google Assistant are also examples of technologies based on machine learning. From Spotfiy, to Netflix, to the new generation of AI chat bots, all of these tools rely on humans themselves to provide the underlying data framework.

 

NATURAL LANGUAGE PROCESSING

Natural language processing (NLP) is the most commonly known and adapted science related to AI and is rapidly advancing thanks to ever-evolving pragmatic business use cases. NLP refers to the way computers analyze, understand, and derive meaning from human language (generally written) in a smart and meaningful way.

Examples of NLP applications to solve computational problems include:

  • Information retrieval – “relevant” information is found when given a query
  • Information extraction – facts are extracted from an audio recording or lengthy text document and stored in a database
  • Speech recognition – a spoken sentence or phrase is recognized and converted to text
  • Question answering – a set of documents is searched to find the right answer after interpreting a question

The most common example of NLP use in daily life is when typing into Google and Google predicts the rest of the phrase. Contract review systems have utilized NLP for years, focusing not only on volume review but also assistance with understanding the content.

 

TECHNOLOGY ASSISTED REVIEW

Both natural language and machine learning techniques are used in technology assisted review (TAR) during e-discovery to speed up the review of gigantic document sets. These systems function primarily on an information retrieval task basis that allows documents to be separated and identified from a larger body of documents based on a query. Beyond the increased speed, a properly trained application will also provide improved review consistency and, in many instances, increased overall accuracy.

Current legal technology applications are shifting from a purely information “retrieval” concept towards “extraction” applications that search and extract useful data that can then be used in reports, predictions, and other tools to help legal professionals make better-informed decisions. These applications now span a number of areas of contract and case law. Contract management departments use NLP and TAR to create reports that summarize terms across contracts and then compare them to established standard terms for variance and risk assessment.

 

DEEP LEARNING

Beyond machine learning is the bigger goal of deep learning, which utilizes more advanced algorithms to perform more abstract tasks. The concept behind deep learning is that it functions much more like a human brain. While there is still an initial human component required, deep learning algorithms have a significant advantage over earlier generations of machine learning in that they make better use of much larger data sets, arguably establishing a path to better, more accurate and more relevant predictions. One example of this concept is Google Translate, which uses an algorithm based on deep learning for translation. There is still much work to be done, and while deep learning has helped to solve some very difficult problems, it remains very far from achieving a human level of understanding.

Ultimately, computer technology is evolving to help machines become better at their tasks with experience. Increased availability of access, along with decreasing costs of AI platforms, is good news for opportunities to leverage smarter and more powerful application platforms today and in the near term.   Google, Amazon, Microsoft, Salesforce, and other companies are making powerful machine learning infrastructures available via the cloud. While we are far from machines that exhibit general intelligence across diverse domains, the increased use of AI can only be expected to gain momentum.