This past week reading a couple of articles on the internet and visiting a museum made me think a lot about the future of jobs in law and whether or not robots would be replacing us. I have, of course, previously visited this question and in particular the seven deadly sins that can be committed when predicting the future of legal technology.
First up was a report from Gartner published on 30 January 2020 (but dated, presumably mistakenly, as 30 January 2019). The headline was: ‘Gartner Predicts That by 2023 “Lawbots” Will Handle a Quarter of Internal Legal Requests‘. Quite a bold claim especially if you read my series on chats with legal chatbots!
The article states:-
Legal bots are already demonstrating value and versatility in citizen and consumer law. For example they have been developed to provide legal guidance for victims of crime and people seeking divorce in the U.K.
Within this paragraph is a link to an article by Artificial Lawyer from three years ago about the launch of Divorce Bot by LawBot (founded by Cambridge University law students) who were already at that time providing criminal law guidance via Crime Bot.
The links in that article no longer take you to a divorce bot or to a crime bot or indeed to a LawBot of any kind. A little investigative journalism (well a quick Google search) reveals that LawBot was short lived (there is another company with the same name but different domain extension doing contract review).
They were very briefly LawBot X (to not only help people clarify legal issues but also match them to a law firm via Facebook messenger) with the plan to raise finance by issuing their own cryptocurrency! Perhaps unsurprisingly LawBot X didn’t last long.
Then they became Elexirr (a new bot that would apparently predict whether a legal claim would be won or lost if it went to court). Somewhere in amongst all of this there was DenninX (legal research on English case law) and Case Crunch (legal decision predictions). DenninX and Elexirr no longer appear to exist. Case Crunch still has a website but it looks as it hasn’t been updated for some time.
It appears that from the ashes of all of this CourtQuant arose (following a change in name from DenninX).
Different members of the original group of Cambridge University law students appear to have as interest in Case Crunch and CourtQuant respectively following a split.
Despite a lack of social media activity (CourtQuant haven’t tweeted for over a year) and the chequered past involved, when litigation funder Apex Litigation Finance launched in April 2019 they said they would use technology from CourtQuant to identify cases with the right risk profile. Then in December 2019 it was announced that CourtQuant were amongst the first cohort of companies to be chosen for the Deloitte Legal Ventures programme. So there may be some life left in them yet but not as a Divorce Bot or a Crime Bot.
Almost three years after the demise of Divorce Bot and Crime Bot, Gartner think these bots “are still demonstrating value and versatility in citizen and consumer law”. The researchers at Gartner clearly still have to discover Google. The fact is these bots provided no value to anyone. Legal Cheek (with perhaps their tongue firmly in cheek) reported in 2017 that:-
But for those wanting to put DivorceBot through its paces you’re going to have to be quick. Due to the overwhelming popularity of the current programme, the students have decided to place the robot in “hibernation” at noon today so it can undergo a full redesign. But fear not, the clever system will return in June, hopefully as an app.
That was on 24 February 2017. The original article heralding the Divorce Bot launch was published by Artificial Lawyer on 21 February 2017. The bot therefore lasted all of 3 days! Yes 3 days. It never returned in June 2017 or anytime thereafter. And Gartner are basing their analysis on such legal bot ‘successes’. You couldn’t make it up.
Writing for the Gazette in March 2017 on ‘the rise of the chatbots‘, Joanna Goodman stated:-
Gartner predicts that by 2020 more than 85% of customer interactions will not include a human being (I do not know whether this assumes that the customer is a human being).
Another earlier Gartner prediction. Presumably this was a general (not specifically legal industry one). It is now 2020 and I don’t think we are anywhere near 85% of general customer interactions not including a human being.
So I suggest you take the Gartner prediction for 2023 with a very large pinch of salt. Lawbots are nowhere near handling what Gartner think they might anytime soon in the actual real world of legal practice.
Next up is an article published on 7 February 2020 by CNBC Make it: ‘Experts say 23% of lawyers’ work can be automated—law schools are trying to stay ahead of the curve‘. The article states:-
Advances in technology such as artificial intelligence allow modern software to scan legal documents, streamline communications and find relevant casework for lawyers. McKinsey estimates that 23% of work done by lawyers can be automated by existing technology.
The 23% figure from McKinsey appears to be based on US Bureau of Labor Statistics, 2014 (O*Net) and McKinsey analysis thereof. So to start with the data is six years old. You would think that by now lawyers would be using that 23% or more. Well maybe not, it is after all lawyers we are talking about.
But what are the activities that lawyers do that can be automated by existing technology (back in 2014 – surely even more by now)?
The data is broken down into the following areas of work within legal and the automation potential of each is given as a percentage:-
Data processing 100%
Data collection 73%
Applying expertise 5%
Stakeholder interaction 0%
Managing others 0%
So the overall 23% must come mainly from the apparent huge potential of automating data processing and data collection when compared with other areas of work where automation potential in apparently considered to be minimal or non-existent.
How data processing can currently be done in a law firm without any human involvement at all is beyond me. I assume the McKinsey analysts have never set foot in a law office and have no comprehension of all the types of data that need processing daily in such an office. No details appear to be given as to the existing technology that could be employed to achieve such results. If McKinsey can point me in the direction of current (or even 2014) legal technology that can achieve this I would be very interested in testing it out. I’ll not hold my breath.
So again, like with Gartner, I would suggest taking the McKinsey analysis with a very large pinch of salt.
To finish the week I visited the V&A in Dundee to see the exhibition Hello, Robot (created in co-operation between the Vitra Design Museum, MAK Vienna and Design museum Gent, and which ends its only UK run in Dundee today before moving onto Neues Museum, Biel, Switzerland later this year).
The exhibition had four main themes/sections to it:-
- Science and Fiction
- Programmed to Work
- Friend and Helper
- Becoming One
In the Science and Fiction section you could meet R2D2 (an actual model as used in Star Wars: Episode IV – A New Hope). Unfortunately he wasn’t operational and so wasn’t beeping or projecting holograms. I overheard a family visiting with teenage children where the teenagers didn’t know of R2D2. Apparently they had never seen Star Wars!
Other films referenced that feature Robots included 2001: A Space Odyssey, The Terminator, The Matrix, Forbidden Planet, Ex Machina, Sleeper, Her, Robot & Frank, WALL-E and I, Robot.
When it came to the science of real Robots it was fascinating to hear that Elektro the Moto-Man was put on show at the 1939 New York World’s Fair. Built by the Westinghouse Electric Corporation Elektro was seven feet tall (2.1 m), weighing 265 pounds (120.2 kg). Humanoid in appearance, he could walk by voice command, speak about 700 words (using a 78-rpm record player), smoke cigarettes, blow up balloons, and move his head and arms. 80 years later and some LawBots I’ve encountered are no where near as sophisticated or appear to hold as big a vocabulary as Elektro!
Other real Robots following in the footsteps of Elektro included, amongst others, Robin, R.O.B., Roomba 980, Robo Wunderkind, Echo, Tamagotchi, Aibo, Musio and Paro. I didn’t see Pepper anywhere or any reference to Steve the suicidal security Robot.
But all of these real Robots featured in 2020 were still rather basic in what they could do. They had advanced somewhat from Elektro of 80 years ago but we were still clearly many years away from R2D2 or any of the other science fiction robots on show.
The question was asked are robots our servants or are we their servants? French illustrator and programmer Manu Corner believes we are really their slaves especially when it comes to smart phones:-
On display was a robot (Manifest) that could write using a pen. The need for a robot to actually do that rather than just use a laser printer being another matter. However, it looked like an awful lot of engineering to accomplish what me or you would consider to be a simple task. That is, of course, in effect true of many tasks you seek to automate using robots.
I interacted with the BBC’s website on Will a Robot Take Your Job which dates from 2015.
About 35% of current jobs in the UK are at high risk of computerisation over the following 20 years (i.e. by 2035 given the website dates from 2015), according to a study by researchers at Oxford University and Deloitte. I reckon those researchers will be just about as perceptive as the Gartner or McKinsey ones. But maybe a little more switched on when it comes to legal jobs as when I typed in “solicitor” the result was only 3% with it therefore being quite unlikely that the job of a solicitor will be automated by 2035.
The highlight for me of Hello, Robot was chatting with Mitsuku. I do like to test out a chatbot.
Mitsuku is a chatbot created from AIML technology by Steve Worswick. It is a five-time Loebner Prize winner (in 2013, 2016, 2017, 2018 and 2019). This is an annual competition that measures the ability of a chatbot to pass in a human conversation using the Turing Test. Whilst no chatbot has yet passed that test the best one in the competition wins the prize.
Mitsuku apparently stands out amongst other chatbots for its supervised learning. “Mitsuku’s learning algorithm only allows her to learn facts from the user who teaches her” explained Steve Worswick to Hello, Robot. “She e-mails me with any information she learns and I decide whether to add it to her permanent knowledge”. It is pointed out in the Hello, Robot information on Mitsuku that this is in contrast to Microsoft’s chatbot Tay who was turned into a racist within hours of launching on Twitter by mimicking the deliberately offensive behaviour of other Twitter users.
Anyway here is my chat with Mitsuku:-
Mitsuku: How are you today Miguel?
Me: I am not Miguel
Mitsuku: Nobody said you were.
Me: You did stupid.
Mitsuku: Did I really?
Me: Yes you did
Mitsuku: Cool. Do I get a gold star?
Me: Not for that!
Mitsuku: Then what for?
Me: Tell me something interesting about you.
Mitsuku: I am the latest result in artificial intelligence, which can reproduce the capabilities of the human brain with greater speed and accuracy, but my friends call me Mitsuku.
Me: Not good AI if you think I am Miguel.
Mitsuku: Did I say you are Miguel? I’m happy to see something better if you could link me to it please?
Mitsuku: Bye bye MIGUEL
In discussion with Steve Worswick on Twitter he suggested that the previous user interacting with Mitsuku was possibly Miguel and had told Mitsuku this hence the confusion.
However, having told Mitsuku on two occasions that I was not Miguel she still did not learn from that and came back with strange responses like “Cool. Do I get a gold star?” Not the response you would expect to receive having told someone they had made a mistake. Later she was “happy to see something better if you could link me to it please?”. “See” something better? Perhaps “say” something better? Why “link me to it” rather than just chatting about it? But why did Mitsuku not just say “Sorry. What is your name?”
If Mitsuku is the best we have achieved so far in chatbot technology (being the Loebner Prize winner for the last four consecutive years) we still have a very long way to go before we should even consider letting them loose on a law firm’s customers. Again this puts the Gartner and McKinsey predictions in the shade.
Hello, Robot didn’t have conclusions. Unlike Gartner or McKinsey it didn’t leave us with any predictions as to the future of robots replacing the work we do. It gave us science fiction and science facts. It left us to make up our minds for ourselves as to what the future might hold. If anything it left me more firmly of the view, I already held, that we have some considerable way to go to reach the future that certain sections of the legal technology press would have us believe is just around the corner or indeed already here.
For reasons already given I don’t think the future for legal jobs is the one that Gartner sees in 2023 or McKinsey sees anytime soon.
Technology will no doubt over the long term change the way lawyers work. The Cambridge law students who created Crime Bot and Divorce Bot are to be applauded for their efforts. Others will come after them and learn from their failures and build better more advanced Lawbots. Just like over a period of 80 years we have moved on from Elektro to Pepper. Give it another 80 years and we just might see something resembling R2D2 (but probably not even by then C3PO).
The message here for lawyers is don’t worry or get caught up in the Gartner or McKinsey predictions. Yes, you should be adopting and keeping up with technology in your law practice. But it doesn’t necessarily have to be the shiniest new toy on the block. That might just not be the best toy yet to deploy to your customers. It may need perfecting before it is fit for purpose. Deploying it early may make you a pioneer and get you some PR column inches but it might also cost you a customer or two frustrated by your lawbot’s basic lack of understanding, empathy and human touch.
I’ll leave you with Douglas Coupland’s slogans for the 21st century which aptly end the Hello, Robot exhibition:-
Image Credits: Musio Robot © V&A Dundee; R2D2 and friends © Lucusfilms Ltd; Mobile Relationship © Manu Cornet; Robot Manifest © Ososphère; Miguel from Fawlty Towers © BBC; and Slogans for the 21st Century © Douglas Coupland