.png)
Artificial Intelligence (AI), a term first introduced by Stanford Professor John McCarthy in 1955, has since become a driving force in modern technology. McCarthy defined AI as “the science and engineering of making intelligent machines”, emphasizing the development of systems capable of performing tasks that typically require human intelligence. These tasks include reasoning, learning, and decision-making based on vast datasets that are beyond the scope of human analysis.
In simpler terms, AI enables computers to learn, make decisions, and perform tasks that resemble human behavior, often at a speed and scale beyond human capability. But how does this actually work? Another question to consider is what are the practical implications in legal practice.
These questions are particularly important when considering how AI is starting to be implemented in the legal sector. Big Law firms have started developing their own AI tools to assist their lawyers in their day-to-day tasks. Not only that, but the future of AI in law is also reflected by the numbers, for example, Goldman Sachs stated in 2023 that 44% of legal tasks can soon be automated.
The present article will provide a quick overview of artificial intelligence, addressing its scientific foundations, and theoretical types, with examples and use cases from everyday life. The article will be part of a series giving insight on not just AI as a field, but in each of the technologies that originate from it. It is important in relation to the widespread use that AI is gaining more and more in the legal industry.
To understand the science behind the field it is important to remember the goal is to build machines capable of performing those tasks autonomously or semi-autonomously without needing explicit human instructions for each step. To accomplish this, AI involves a range of technologies that allow computers to carry out complex tasks, such as recognizing visuals, interpreting, translating speech and text, analyzing data, providing recommendations, and more.
Before delving into the inner workings of AI, there is a need to conceptualize a few terms, such as algorithm and model. An algorithm is a set of instructions designed to perform a task. When it comes to AI, an algorithm is applied over time and it is what enables a machine to learn how to perform a task on its own regardless of human intervention. The set of instructions controls what functions are performed on the data supplied to the algorithm, and they are also relevant to the task the algorithm is intended to perform. A model is the product of the algorithm after it analyses the supplied data.
AI works in two ways: input and output. The first approach to AI involves inputting data relevant to a specific task or goal, where the relationships between variables in the data are already known, into an algorithm. The algorithm processes this data to create a model, essentially a mathematical and statistical representation of patterns within the data. This model can then be used to generalize and predict outcomes. Simply put, AI works by using the data you have to generate insights or predictions about data you do not have. The strength of this approach lies in its ability to produce valuable output based on large quantities of high-quality, well-structured data, helping to make informed decisions about unknown or hypothetical scenarios.
The second method, called reinforcement learning, mirrors the way humans learn most tasks - through experience and feedback. Instead of programming the machine for every possible action, the system learns by interacting with its environment and receiving feedback through trial and error. Actions that bring the machine closer to its goal are rewarded, while actions that do not are penalized. This approach is particularly useful for complex tasks, such as robotics, where programming for every possible scenario is impractical, and learning through experience becomes a more efficient solution. An example of reinforcement learning can is the robot that learns through trial and error. Such robots have also been built by universities, such as Berkeley and Leeds in the UK.
.png)
AI also uses indirectly other non-technological disciplines, such as linguistics, neuroscience, philosophy, and psychology, which work with computer science, data analytics, and engineering to train machines to mimic human cognitive processes. The backbone for the training of the machines is exposure to large amounts of data, which allows AI systems to “learn”, improve, and make better decisions over time. Comprehending these foundations assists in explaining why AI is a powerful tool that transforms industries and everyday life.
AI comes in many forms and at times it is difficult to distinguish whether a tool really uses AI, hence it is important to understand the general types that exist. The easiest way to comprehend is to consider them as more like theories of AI types and not applications of AI models.
Reactive machines: These AI systems are limited, as they only react to different kinds of stimuli based on programmed rules. They are limited, also because they have no memory and they are task-specific. An example of a reactive machine is IBM’s Deep Blue a task-specific model that specializes in playing chess, it is the machine which beat chess grandmaster Garry Kasparov in the 1990s. The system could identify pieces in the chessboard and make predictions, but because it had no memory it could not inform future matches.
Limited memory: These systems go a step further, as the name suggests, as they have limited memory, hence they can inform future decisions using past activity. An example of this limited memory system is a part of the functions of self-driving cars.
Theory of mind: This is a psychology term and when applied to AI it refers to a system that can comprehend emotions, such a system does not exist yet. However, research is evolving on the possibility of materialization.
Self-awareness: This goes a step beyond the theory of mind, and at this point, we can understand that the four types could be placed on a scale regarding their functions and possibilities since some are more advanced than others. A self-aware AI is a machine that is aware of its existence and has both the intellectual and emotional capabilities of a human. It also does not exist like the theory of mind.
Another way to separate the categories of AI is by what they do.
Narrow AI refers to systems designed to handle specific tasks, operating within defined boundaries without the ability to generalize beyond their programmed functions. Examples include virtual assistants like Siri or Alexa and recommendation algorithms used by services or Netflix.
General AI (or AGI) is a theoretical concept where machines can perform any intellectual task a human can. AGI would require the ability to reason across diverse domains and solve problems without predefined instructions, like leveraging fuzzy logic to navigate uncertainties rather than relying on binary decision-making.
In upcoming articles, we will further explore the various subsets of AI introduced here, offering a more detailed examination of their roles and applications:
One of the most groundbreaking advancements in Legal tech is Generative AI. The idea of AI being used in the legal sector became popular in the 2020s but the application of Generative AI having a true impact by firms using it in their everyday tasks became a reality in 2023. However, it is not the first form of AI that lawyers have used. Generative AI enables systems to create new content - whether it is drafting or summarizing documents - based on learned patterns from large datasets. LLMs exactly like casepal are examples of how AI can assist with tasks that once required significant human effort.
Artificial Intelligence seems as a quite recent topic, however, it has been around for many years in different forms. Everyone has encountered AI in their life without realizing it before platforms such as ChatGTP or specialized AI tools like casepal came. Lawyers have been using AI not in the form we know it nowadays but in different forms.
An example of an AI function is email management. So, when email filters professional or personal inboxes and automatically sorts incoming messages in the folders “Important” and “Spam”, it actually uses AI to help users prioritize important communications. It specifically relies on machine learning algorithms that analyze the content, sender, and patterns of past interactions to prioritize your communications.
Another example lawyers often encounter is spell check and grammar correction tools, such as Grammarly or the built-in spell check Microsoft Word has to automatically detect or correct spelling and grammatical errors. These systems have been around for years and use natural language processing to understand context and provide better suggestions.
Legal research platforms, such as Westlaw also incorporate AI to enhance search functionalities. These platforms use natural language processing (NLP) to understand queries more intuitively, helping lawyers find case law, statutes, or legal precedents more efficiently by identifying relevant results based on context rather than just keywords.
Another example is voice assistants like Siri or Google Assistant, which lawyers might use to schedule meetings, set reminders, or dictate notes, also rely on AI through voice recognition and natural language understanding.
Finally, facial recognition is also an everyday example of AI as it is found in devices like smartphones and laptops. It allows people to unlock their devices securely by scanning their faces. This AI-powered feature analyzes facial features and matches them to stored data, providing a fast and secure way to access information.
AI is evolving rapidly and is getting integrated into various aspects of our daily lives. It is reshaping the way we interact with technology. While the full potential, including self-aware machines, remains theoretical, its current applications are making significant strides in enhancing efficiency in everyday tasks, especially on a professional level. This is greatly significant for lawyers as we have come to realize in the last years, as evidenced by the trust firms have put in AI tools. The extent of AI use will only grow and transform the legal industry, with more professionals in the field wanting to be in the know.
The opportunities that Generative AI provides have been already recognized by big law firms, as indicated by the process of them building their own AI platforms for their employees. This is coherent with the overall idea that Generative AI, and particularly specialized AI tools powered by vertical LLMs such as casepal are having a great impact on making professionals more effective and efficient.
As a last point, it is essential to understand the basics of how Artificial Intelligence systems function, and the vast possibilities they hold for the future.
To that end, the next article’s topic explores key subsets of AI, such as machine learning, natural language processing, robotics, and generative AI, each contributing to the diverse applications and capabilities of AI today.
TechTarget, ''AI (Artificial Intelligence)'' https://www.techtarget.com/searchenterpriseai/definition/AI-Artificial-Intelligence accessed 11 October 2024.
Google Cloud, ''What is Artificial Intelligence?'' https://cloud.google.com/learn/what-is-artificial-intelligence?hl=en#how-does-ai-work accessed 11 October 2024.
MIT Technology Review, ''What is Artificial Intelligence? The definitive guide'' https://www.technologyreview.com/2024/07/10/1094475/what-is-artificial-intelligence-ai-definitive-guide/accessed 11 October 2024.
Stanford University Human-Centered AI Institute, ''AI Definitions''
https://hai.stanford.edu/sites/default/files/2020-09/AI-Definitions-HAI.pdf
accessed 11 October 2024.
IBM, ''AI Model'' https://www.ibm.com/topics/ai-model accessed 11 October 2024.
IBM, ''Artificial Intelligence'' https://www.ibm.com/topics/artificial-intelligence accessed 11 October 2024.
Kasia Borowska, ''The What, How, and Why of AI'' https://www.forbes.com/sites/kasiaborowska/2020/11/03/he-what-how-and-why-of-ai--part-3/ accessed 11 October 2024
Thomson Reuters, 'Introducing AI-Assisted Research: Legal research meets Generative AI' (Thomson Reuters, 6 September 2023) https://legal.thomsonreuters.com/blog/legal-research-meets-generative-ai/ accessed 19 October 2024.
GeeksforGeeks, 'Face Recognition Using Artificial Intelligence' (GeeksforGeeks, 2 December 2020) https://www.geeksforgeeks.org/face-recognition-using-artificial-intelligence/ accessed 19 October 2024.
TechTarget, 'What is Face Detection?' (TechTarget, 19 January 2022) https://www.techtarget.com/searchenterpriseai/definition/face-detection accessed 19 October 2024.
Will Louch, ‘Inside Linklaters’ AI Push: "It Feels Like a Tech Company"’ Financial News (London, 8 November 2023) https://www.fnlondon.com/articles/inside-linklaters-ai-push-it-feels-like-a-tech-company-20231108 accessed 28 October 2024.
Andrew Maloney, ‘Forget ChatGPT: Law Firms Are Launching Their Own Gen AI Chatbots’ Law.com (7 September 2023) https://www.law.com/2023/09/07/forget-chatgpt-law-firms-are-launching-their-own-gen-ai-chatbots/ accessed 28 October 2024.
Konstantinos Chatzilygeroudis, Vassilis Vassiliades, and Jean-Baptiste Mouret, ‘Reset-Free Trial-and-Error Learning for Robot Damage Recovery’ (2018) Robotics and Autonomous Systems https://www.sciencedirect.com/science/article/pii/S0921889017302440 accessed 28 October 2024.
Public Affairs, ‘New Deep Learning Technique Enables Robot Mastery of Skills via Trial and Error’ Berkeley News (Berkeley, 27 May 2015) https://vcresearch.berkeley.edu/news/new-deep-learning-technique-enables-robot-mastery-skills-trial-and-error accessed 28 October 2024.
University of Leeds, ‘Worker Robots That Learn from Mistakes’ University of Leeds News (Leeds, 12 January 2017) https://www.leeds.ac.uk/news-technology/news/article/4494/worker-robots-that-learn-from-mistakes accessed 28 October 2024.
Patrick Palimore, 'Clio’s Legal Trends Report 2024: AI, Automation, and the Death of the Billable Hour' (TechLaw Crossroads, 11 October 2024) https://www.techlawcrossroads.com/2024/10/clios-legal-trends-report-2024-ai-automation-and-the-death-of-the-billable-hour/ accessed 29 October 2024.
PwC, 'Are You Using AI Yet? How Lawyers Can Use It to Their Advantage' (PwC, October 2024)https://www.pwc.com/gx/en/about/pwc-asia-pacific/are-you-using-ai-yet-how-lawyers-can-use-it-to-their-advantage.html accessed 29 October 2024.