We all live in a time of electronic marvels. From dancing robots to space exploration, we are constantly in awe of the technological progress of the past 50 years. Without a doubt, digitization and cloud computing has made technologies that were only available at high costs accessible to practically anyone with a simple mobile device.
It’s no wonder that the world is fascinated with the marvels and promises of Artificial Intelligence (AI). It almost seems like computers are reaching a level close to human intelligence. Since the early “Twilight Zone” black and white series of the 1960s, we have speculated about the possibility of intelligent interactions with AI-powered machines. Perhaps the time has come?
But more to the point – how will AI affect the practice of law? In terms of business and geopolitical concerns, how does AI introduce new business opportunities and efficiencies for companies and law firms? For this, we need to first dispel the many myths of AI and honestly acknowledge the many unknowns about AI.
What is AI?
Conceptually, AI started to come up in academic journals in the earliest days of computing. For the vast majority of the 20th century, that’s where it stayed. Instead, simpler conceptual approaches rooted in mathematical regression analysis and traditional machine learning approaches have achieved more wide-spread and generalized application.
In 2012, this all changed. Prof. Geoffrey Hinton of the University of Toronto, leading a team of graduate students, demonstrated breakthrough results in an image recognition challenge. For the first time, and in a dramatic fashion, they demonstrated that AI models can achieve far better results than traditional machine learning approaches. With a big bang, the gold rush of AI began.
Why is AI different from machine learning? How?
AI is fundamentally different from traditional machine learning because of the incorporation of neuron-like computing units that are loosely modelled on human neuron cells. However, that alone is not enough to achieve productive results. Together with modern massively-scalable cloud infrastructure, the availability of very large data sets, and the modern techniques to optimize AI models with billions of coefficients, AI demonstrates that it can achieve far more impressive results than traditional machine learning approaches.
To train or not to train – What does training AI models mean? Who does this?
When people learn about AI and the training of AI models and the inherent complexities with evaluating mathematically complex AI models, the question always revolves around talent availability. Who has the expertise to train new models? Do we expect attorneys to do this? What is even required to train a model?
The truth is that AI model training is indeed a specialized skill although there are many new approaches that are starting to make AI model development more accessible to the masses. However, to effectively build models that are broadly useful, it often requires enormous multi-day computing runtimes and AI expertise that are typically not accessible to even some of the most advanced companies. Without a doubt, most attorneys are already overbooked – getting attorneys to help train models is a non-starter.
Are AI attorneys better than real attorneys? Or are they just different?
We need to be careful about what we can reliably achieve through “AI Attorneys”. At My Legal Einstein, we firmly believe that the proper place for AI is to assist attorneys in their roles – to reduce the monotony, to find information more readily, to help navigate complexity, to achieve improved efficiency. We simply do not believe that an AI system should give legal advice. After all, even the best AI is NOT capable of the sophisticated thinking necessary to formulate and distill a legal opinion. Simply put, with today’s technology, AI Attorneys are not viable. Instead, we should focus on how AI helps in terms of productivity and efficiency – in other words, in terms of augmenting the attorney’s intelligence. Instead of thinking about “AI Attorneys”, we should think about AI-powered Attorneys – Attorneys who can successfully leverage the capabilities of AI to achieve their own breakthrough results.
How do I know AI works? How do I know it doesn’t work?
AI works when it saves the attorney time and makes them more productive. This is the standard that should be applied to all technology applications. We lose sight of the fact that AI is not intended to solve all problems and it’s not an answer upon itself. For AI to be successful, it’s impact should be immediate and the measurable value should be obvious.
Often, “AI solutions” are positioned as novelty products, making recommendations as an “AI Attorney” but failing terribly to make a positive impact. The reason is simple. When AI makes a recommendation, the attorney often needs to double check that it’s a valid recommendation. Then, the attorney needs to check if there were recommendations that were missed. Worst yet, after the attorney reviews and corrects the recommendations, the next time AI makes recommendations, it repeats the same errors over and over again. In a sense, when AI is implemented poorly, it behaves much like a paralegal with poor training, creating more issues than they solve.
Does it make sense to implement AI in my organization?
There are many reasons to adopt AI-powered solutions in your organization. However, it’s important to understand the limitations of the technology and to adopt AI to solve specific problems with measurable productivity gains. With this in mind, companies should consider the following guidelines on how best to get started with AI to achieve success.
- Do not expect attorneys to train AI models – there’s no time for that. Leave it to the solution providers.
- Look for AI-powered solutions that are “pre-trained”, so that you can achieve fast time-to-value. Projects should realize benefits immediately, not after a long costly implementation.
- Don’t fall for using AI technology for technology’s sake. It’s always tempting to jump on the latest trends.
- Most importantly: Try, experiment, and measure – Compare your current approach to using an AI-powered approach and honestly measure the performance and productivity gains. Did the AI-powered approach save you time, improve efficiency and deliver better results?