By Jim Chiang.
We’ve all heard the stories of how the intelligent AI agents will come and rule the earth. For attorneys, the worst fears are realized as sensational media stories depict a future in which the lawyers are replaced with legal drones, making legal decisions on our behalf. Fantasy or reality? We often mistake the sci-fi of the movies as real projections of the future. We need to separate reality from fiction to understand how AI, as the case with any new technology, can be leveraged to help us and not hurt us.
Origins of Artificial Intelligence
Artificial Intelligence was officially established in academic circles starting in 1956. As a simple history lesson, this was 3 years before the first commercially viable transistor was invented. Indeed, it was at the very beginning of computing that dreamers started imagining a world in which computing intelligence would start performing at the levels of human-level intelligence.
AI - What took so long? Fast forward 65 years and you can start seeing some of the great AI inventions of our day … semi-autonomous driving, dancing robots, Siri/Alexa talking devices - the list goes on and on. But how did we get here. Those in the academic world would tell you that there have been several periods of AI winters … periods of optimism followed by disappointment as the next invention was just a breakthrough away. It was always next year that AI was going to be the next big thing … and this happened for many, many years.
Modern AI breakthrough
Then, one day in 2012, a couple of graduate students at University of Toronto under the direction of Prof. Geoffrey Hinton, one of the few professors who continued to work on AI, published a ground-breaking paper that demonstrated how AI can be applied for breakthrough results on an image recognition computing challenge. Here’s the link. Practically overnight, the AI gold rush started. How did they do it? They combined three critical elements - Deep AI neural models, huge computing power, and large quantities of data. It turns out that the key elements of AI were only made readily available by the modern advances in cloud computing and large scale computing infrastructure.
AI vs. Traditional Machine Learning
But isn’t AI just another form of Machine Learning? Isn’t it the same thing? In a word, no. Machine learning has been around much longer in very practical applications, especially related to Regression Analysis. AI is fundamentally very different with much more complexity of models and the number of parameters to tune to make AI work effectively. AI also uses “neural” models which are modelled similarly to human neurons. However, AI should not be considered “better” machine learning, as AI only applies when there are huge amounts of data resources - like images, sound, and text.
Will AI transform the Legal practice and how?
It’s never been a question of “if”, it’s always a question of “when” and “how”. AI will transform the legal practice in many ways; however, we are still very early in the technology adoption cycle. As a technology vendor, we are always developing new methods that introduce AI in a positive and productive experience so that end users can experience the benefits. However, we need to be very careful of introducing technology that is simply not ready … i.e. the cost of implementing an algorithm may be higher than the realized benefits. High tech has always suffered from the “hype” machine - always painting the picture of what it could be but not what it is. As vendors, we need to make sure that we help customers realize real benefits as opposed to adopting AI for technology’s sake.
Can AI be trusted?
Can AI be trusted? Well, it depends (purposely borrowing lawyerly language). One of the biggest applications of AI and machine learning in general is in the application of targeted advertisements and risk predictions (for loan applications, etc.). When AI is used for decisions on humans, we need to ensure that AI is free of bias implicit from the training set, so that AI-based decisions on humans do not reinforce bias. Certainly, facial recognition AI technologies should be regulated as it infringes on privacy concerns. For the legal industry, AI solutions vendors need to continue to architect AI in a manner that does not make recommendations to an attorney but instead, help the attorney navigate and find information that is critical to their work.
Doesn’t the billable hour limit AI innovation?
Yes and no. The reality is that more often than not, legal tech vendors point to the billable hour as the reason why technology adoption is slow. Although it could be a contributing factor, we see a lot of pressure from law firms to continue to demonstrate more value per billable hour. There’s a lot of momentum from some of the larger law firms to modernize and demonstrate technology leadership to their clients. In fact, many law firms are already adopting AI in specialized applications that they highlight to their customers and that introduce new legal service offerings that were previously too labor-intensive.
How to get started?
As with anything, the adoption of legal technology and artificial intelligence is a journey of experimentation, education, and adoption. For this reason, it’s critical for AI legal tech vendors to minimize the upfront investment needed to experiment and educate on AI-based solutions. After all, there is very little bandwidth from attorneys to become AI engineers.
About the Author
Jim Chiang is the CEO and Chief AI Nerd for My Legal Einstein, an AI-powered Contract Execution Platform for the review, collaboration, negotiation, and execution of legal contracts. Jim previously led the AI engineering efforts at Conga/Apttus and Icertis and has over 20 years of experience in big data management and AI algorithm development.