23 Nov 2016

An AI Trained By Lawyers

A new start-up called Luminance is being heralded as a new breakthrough in the development of AI, specifically designed for the legal sector. Focussing on assisting with the reading of huge amounts of data, it is quickly establishing a reputation as a game-changer.


Luminance in the Legal World

One of the key areas in law that an AI can be put to use is the repetitive, manual work of reviewing documents, particularly in the cases of acquisitions and mergers. Tackling the mountain of data – which is exponentially larger in the digital world — is a process prone to human error, simply because of the sheer size of data involved.

As Emily Foges, CEO of Luminance, puts it, “One of the least gratifying tasks a lawyer performs is reviewing documents, a process often left to junior associates. An average data room contains 34,000 pages, for example, making the task gargantuan and prone to error.”

Luminance was founded at the University of Cambridge and is run by people from Autonomy some of whom, like Robert Webb (ex-GC at British Airways), are steeped in law and legal process. Backed by technology investment fund Invoke Capital, it is very much part of the trend of law firms to employ all technological means possible to improve workflow and efficiency. “AI is fast entering the mainstream. Luminance was launched one month ago, and the response has been phenomenal, with over 100 firms contacting Luminance asking for a demo,” says Foges. 

Smart Algorithms

The advantage that Luminance holds over other legal AI systems is that it doesn’t rely on contextual or key word searches. “These require a lawyer to know what he or she is looking for,” says Foges, “whereas Luminance surfaces the "unknown unknowns” – contracts in unexpected jurisdictions, missing pages, etc.”

Instead of being a rules-based system or one based on searching for keywords, the algorithms that power Luminance are designed to make it learn. 

The rules or signature-based approach has inherent weaknesses in that it requires an exhaustive list of the rules required which require maintenance and upkeep; it will have difficulty picking up anomalies such as unfamiliar words or even misspelt words; and improving such a system involves improving the rules, again at a great cost of time and effort, not to mention that there is a limit to the complexity it can handle.

The keyword approach is not only time-consuming, it also relies on the human outside of the system to manually enter in the key terms, expecting them to know exactly what they are searching for. This approach is unable to assist in the search for an unusually phrased clause, or even if words or a whole page are simply missing.

Luminance is built to learn, not only from the input of others, but by teaching itself. The machine-based approach means that the AI can be left to its own devices, becoming even smarter as it analyses more and more data. The algorithms detect patterns in the language to infer understanding of the meaning, bringing its intelligence much more in line with that of a human, but able to read huge numbers of documents in half the time.

Once it is set to work and analysing a large enough data set, it can even detect anomalies. It establishes the norm from reading the documents and can then see if any data points differ from it. And if a clause is considered by the human lawyer to be low risk or not worth flagging up, Luminance can be instructed to ignore those types of clauses to keep the workflow streamlined.

The direct asset that Luminance provides for this type of work is speed. As Foges says, “Lawyers who have used Luminance cite the speed, accuracy and workflow of the platform as benefits of the software.”

Impact

Primarily, the focus of any legal AI system, Luminance included, is to allow the human lawyer to focus their time and effort on those tasks that are higher-up than a computer’s capabilities. As law is considered a cognitive sector, implementation of AI is seen as being a factor that can complement lawyers, rather than replace them.

“In time, it may be that fewer lawyers are needed to review information, but we do not yet see a situation in which we dispense with humans entirely and rely on machines for legal judgements,” says Foges.  

Lawyers that are less dependent on their superiors for supervision and advice are more effective lawyers. Developing AI to be more sophisticated is a transformative step in changing the landscape, seen as being something more than simply improving on old systems in the way Electronically Stored Information (ESI) did. Changing the way that it works holds more significance.

Even though it was done on a computer, the method of processing the information involved in an acquisition still resembled the way it has been done for hundreds of years, while Luminance, and future systems, is intent on changing that process, rather than just speeding it up.

And in the future? Emily Foges says, “As the algorithms improve, and lawyers and their clients become more comfortable with systems like Luminance, we foresee a move into compliance, and General Counsel everywhere adopting these technologies for speed and accuracy.”


related topics