for website

The dawn (and the imminent explosion) of AI in law and legal services

By Steven Malyj, Trainee Solicitor at Riverview Law

Towards the end of last year I watched a TEDSalon talk delivered by Kenneth Cukier to a Berlin audience in June 2014.  During the talk Cukier referred to machine learning and the advent of the self-driving car – how the approach to this was evolving from the challenge of explaining to a machine how to drive (think Hugh Jackman teaching the robot how to box in Real Steel), to a state where we provide the facts and the machine teaches itself (Cukier 2014).  It seemed like a fanciful notion and as I listened I imagined hauliers, taxi drivers and white van men going about their day blissfully unaware of the pending implications that this technology may have on their future.  Surely, now that the UK Government has unveiled its plans to invest almost £19 million in the road legal testing of driverless cars between now and 2020, these workers will begin to fret for their livelihoods.  Then, in January 2015 Riverview Law announced a Knowledge Transfer Partnership (‘KTP’) with the University of Liverpool to (among other things) draw on a pool of skilled data scientists and ground breaking Artificial Intelligence (‘AI’) with the aim of automating some of the cognitive abilities of knowledge workers and provide organisations with intelligent decision support tools. In a momentary panic, the palpitations began, my palms became clammy and I started to picture myself cueing up with aforementioned motorists to ‘sign on’.  But what does it really mean?  Where will AI really take the workforce?  This article investigates the dawn (and the imminent explosion) of AI in law and legal services and dispels the myth that new technology will signify the end for lawyers; positing rather that this is merely a new beginning, bringing with it a host of new opportunities for myself and my learned friends.

I think we can take it as read by now that law firms, traditionally, are resistant to change; or that is at least how it reads on the face of things.  Some of us (myself included) may think it is more the case that law firms are not resistant, they are incapableOf the 304 respondents to the Altman Weil Law Firms in Transition survey of 2014 (Altman Weil, Inc 2014) the overwhelming consensus opined that the legal market had changed in fundamental ways, 67% of them predicted that the rate of change will only increase, but only 13% expressed any confidence in their firm’s ability to change or to at least keep pace with changes (Georgetown Law 2015).  With that in mind, it is perhaps no surprise that technology has not permeated ‘our world’ with any real pace, so before the critics among us nonchalantly snub the future of legal technology as some sort of urban legend – a bed time story about jurisprudential robots that will sneak into our office and steal our desks – it may be worthwhile briefly examining the application of AI in other fields.

It should be enough to refer you to the household names of Amazon and Netflix and the way they use machine learning to suggest books and movies that you may be interested in; or how Facebook and LinkedIn can suggest people that you might know or wish to connect with.  No?  Okay, in October 2012, Geoffrey Hinton and a team of 4 doctoral students won a competition aimed at designing automated drug detection systems.  These systems use algorithms to analyse data sets and identify particular molecules that will bind to targets and therefore make for effective treatments. Interestingly, the competition ran for 60 days, but Hinton and his team did not enter the competition until the last two weeks.  Even more interestingly, neither Hinton nor his compadres had any background whatsoever in Chemistry, Biology or Life Sciences.  They even beat off competition from all of the algorithms developed by Merck – one of the largest pharmaceutical companies in the world.  Hinton and his team achieved this accolade using Deep Learning; a process that attempts to mimic the neural connections of the human brain, processing data in much the same way.  I firmly believe that this is an amazing fete of technology and a sure sign of things to come – law is incredibly complex in its own right, but I defy anybody not to agree that it is at least paralleled by modern medicine and if 5 people with no subject matter expertise can outperform industry leaders then why can it not happen in law?  Make no mistake, machine learning and AI will go everywhere and it will happen at an alarming rate from herein.

Applications of machine learning are certainly serving to further the principles of Moore’s Law – the theory that over time the pace of technological advances will increase on an exponential curve.  Figure 1[1] shows how Moore’s Law is progressing.

Accelerating pace of change

The graph shows that right about now, the processing power of modern technology has surpassed the processing power of a mouse’s brain and that by circa 2025 it will surpass the processing power of the human brain. Now, Moore’s Law isn’t ‘real law’.  It’s not grounded in hard science and empirical facts.  Moore’s Law was a simple observation by Gordon Moore in 1965 that the number of transistors per square inch on integrated circuits roughly doubled each year since the invention of the integrated circuit and predicted that the trend would continue this way.  Looking at the historical plots on the chart, it was quite the prediction, so can we really say that more is not to come?  And quickly at that?

What, then, for law?  The use of AI in law is not really a new concept; the first International Conference on AI and Law was held in Boston on 27th – 29th May 1987 and it has been held on every odd numbered year since then.  In 1988 Philip Capper and Richard Susskind didn’t just co-author Latent Damage Law: The Expert System, they developed a piece of software that could guide lawyers through “a dense web of barely intelligible interrelated rules” (Susskind 2014). The system mapped over 2 million paths that would guide lawyers to an informed solution and it was all extrapolated from the mind of one expert lawyer; it was so good that it was considered better than Capper and it was able to reduce the time that it took a lawyer to do this work from 10 hours to just 10 minutes.  But this was a very niche field and so perhaps it is understandable as to why the impact on the legal sector wasn’t what we would call earth shattering.  Further, to extrapolate that technique across the whole of the industry would take a great deal of initiative and inclination to gather the best minds in each field of law for as long as it takes to transfer their knowledge to a system.  We need systems that can learn, spot patterns, understand language and context and adapt.

Germany’s Federal Government have recently deployed an AI application which makes automated decisions regarding citizens’ claims for child benefit.  The system is able to assess claims made by citizens by interpreting the justifications that they put forward in their application and determining whether these justifications merit the award of the benefit(s) claimed (Hodson 2013).  I cannot attest to the grounds for awarding child benefit in Germany or the means and criteria by which such claims are assessed, but I suspect that it is safe to say, like in the UK, that such claims are not straightforward and will include the opportunity for the claimant to provide details of extenuating circumstances and any other relevant information to support their claim and so on. What we have here is an example of an advancement in technology from Capper and Susskind’s method of man teaches machine – machine guides skilled operator, to an instance where man tells the machine the basics – machine learns, understands and operates itself.  The cynic here would shout at the top of their lungs “this will cost real people their jobs”. The optimist (particularly from a UK perspective) might well counter the cynic by pointing to the time it takes to process claims, deal with challenges and respond to the ever increasing backlog of queries – it is fair to say that government organisations are reticent to increase staffing just because the number of claims increases; so, if we take the first step of a claim out of the hands of the human workforce, that workforce is freed up to spend more time quality assuring, assessing challenges and queries and providing the sound customer service that the clients are so often crying out for.

In Helsinki, TrademarkNow is a relatively new start up firm offering a web based trademark management platform that can provide instant trademark search results in as quick as 15 seconds (shaving days off the traditional trademark search methods), provide automated and relevant watch alerts and allow users wider watch capabilities (TrademarkNow 2015).  If you consider the amount of data in existence with regard to trademark and intellectual property – all those companies, all those trademarks, logos and brands – is it any wonder that this area of law has been among the first to see the exploitation of automated systems?  The standard approach to conducting a trademark search is a detailed review of the proposed brand, potential names and whether any other companies exist.  But lawyers will appreciate that it doesn’t end there – we need to consider whether there is any similarity with other brands; after all our clients don’t want to potentially face a claim for passing off simply by virtue of the fact that their customers might mistake the product for another brand altogether.  This is where TrademarkNow gets interesting, because it will go beyond the generic search, beyond even a search for similar names and brands based on phonetics, colour or make up; in the event of a potential match, Trademark now uses AI to review that particular match’s history of defending its brand, producing a risk profile and an analysis of the likelihood that the proposed brand will make it onto the trademark register.

Lex Machina is a Stanford based company that offers data analytics tools to lawyers in the field of IP litigation.  Consider a typical approach to litigation – the client comes to you with a claim that another party has infringed their intellectual property rights in whatever product or service they offer to their customers (or, similarly, that another party has accused your client of infringing their intellectual property rights).  You take their information, you use your understanding of commercial law and intellectual property interspersed with research from a handful of cases and you assess the merits of their claim or defence.  Lex Machina toss aside the merits of the claim.  Instead they use data analytics to analyse the behaviour of opposition, counsel and judges across a range factors to provide a forecast result.  Put simply, it has the potential to accurately predict the outcome of a case with a much greater degree of certainty than any human, because it has the time; it has the inclination (that is to say as far as it is possible for a machine to possess such a trait); and it has the ability to analyse huge volumes of data that a lawyer just would not have the time to conduct (unless of course the client is happy to pay them their hourly rate – and then some – to do so).  It has quite aptly been dubbed “Moneyball for Lawyers” (Doherty 2014).

I feel that it is necessary again to come back to the worry that AI, in practice, will spell cuts for lawyers.  To the contrary, Anna Ronkainen (the co-founder of TrademarkNow) disagrees with any assertion that AI is only going to add a fifth horseman to the Apocalypse – I’m thinking Lady Justice bursting through the flames on a mighty chariot and striking down lawyers left, right and centre.  Instead, Ronkainen suggests that AI will allow lawyers to focus on more meaningful tasks by eliminating manual work, which brings us back to the point of lowering the cost of the service to clients while at the same time increasing its value (Ronkainen 2015).  She further notes how the UK “is at the global forefront of this development with professional liberalisation initiatives such as Alternative Business Structures…” (Ronkainen 2015).

I am largely inclined to agree with Ronkainen.  More to the point, I continue to be excited by the ever changing face of legal technology and I am privileged to work with an organisation that is not only fully prepared to embrace these changes, but is also exploiting the potential that it holds for lawyers and making a conscious effort to dispel the fears that it will reduce the workforce.  To the cynics and the optimists alike, we can only say “watch this space”.

Bibliography

Altman Weil, Inc. Law Firms in Transition: An Altman Weil Flash Survey. Prods. Thomas S Clay, & Eric A Seeger. Pennsylvania, 2014.

Big Data is Better Data. Directed by TED Salon. Performed by Kenneth Cukier. 2014.

Doherty, Sean. “Moneyball for Lawyers.” Law Technology News, April 2014.

Georgetown Law. 2015 Report on the State of the Legal Market. Georgetown Law, 2015.

Hodson, Hal. “AI Gets Involved with the Law.” New Scientist, 2013: 20.

Ronkainen, Anna. “Don’t Rage Against the Machine.” Intellectual Property Magazine, March 2015.

Future of Artificial Intelligence and Law. Performed by Richard Susskind. ReInvent Law, 2014.

TrademarkNow. TrademarkNow. May 2015. https://www.trademarknow.com/ (accessed May 2015).

[1] http://www.reddit.com/r/singularity/comments/2xu2sx/moores_law_2015_mouse_brain_has_been_reached/

Steven Malyj

About the Author: Steven Malyj

http://www.riverviewlaw.com