ChatGPT – Game Changer for Knowledge Management
By now most would have heard of how OpenAI’s ChatGPT has been taking the world by storm since its launch in late November 2022.
Why is ChatGPT prompting so much discussion? Its interface is user-friendly, even for the Luddite. One merely needs to type in any question in the text input box and it will generate articulate and detailed responses in plain English.
It can write code, poems and even pass certain exams. And yes, it can generate coherent contracts and answer legal queries in a thorough and convincing manner.
Unlike Bing which is reportedly “unnerving” and “unsettling”, ChatGPT has in place guard rails. It refrains from commentary or opinion and even has legal disclaimers.
ChatGPT is a major milestone in generative artificial intelligence (AI). The technology can be harnessed and translated into significant practical uses for legal professionals, in particular knowledge management.
This article explores the legal applications that can benefit from generative AI (technology behind ChatGPT), the ethical and security ChatGPT poses and why ChatGPT will not replace lawyers in the near future.
What is OpenAI
OpenAI is an AI research company founded in San Francisco in 2015 and believed to be backed by the who’s who of technology. Tech giant Microsoft, an original investor, recently announced a further multiyear, multibillion-dollar investment in OpenAI.
ChatGPT / Chatbots Explained
ChatGPT is an interactive chatbot from OpenAI that uses generative AI to answer questions on anything posed in plain English. Chatbots are AI systems that use natural language processing (NLP) to understand and respond to human communication.
NLP refers to the branch of AI that deals with giving computers the ability to understand text in much the same way human beings can.
Chatbots are generally constructed under two forms, retrieval and generative. A retrieval based chatbot’s responses are limited to pre-defined responses while the generative chatbot can generate new text rather than select from pre-defined responses.
ChatGPT is an example of generative AI. It can understand the context of a search query and provide results that are relevant and specific to the user’s needs with NLP techniques. It can also interpret the intent behind a user’s query, even if it is phrased in a casual or conversational way. This makes it easy for users to ask questions and get the information they need.
Using ChatGPT – User Beware / Risk management
Lawyers contemplating experimenting with ChatGPT in their work need to be aware of security, client confidentiality and copyright issues. These issues can arise through the transmission of data when using ChatGPT because the security of what is posted into the ChatGPT prompt box is not clear. It is not known how and where those prompts will be stored and whether the prompt will allow the AI to learn your confidential information and then provide it to others.
As with other legal research platforms, publications or legal software, ChatGPT is just a tool or service to be used. A lawyer’s ethical obligations must always take precedence over convenience. The final review of work and legal advice still lies with the lawyer.
ChatGPT should be explored with caution. Start with controlled tasks that do not require input of sensitive information and be sure to do adequate quality reviews of the responses.
ChatGPT’s Potential as a Knowledge Management / Legal Research Tool
Generative AI has the potential to improve efficiency and effectiveness of supporting tasks that are time consuming and mundane, such as extracting structured data from an unstructured database.
The current ChatGPT prototype shows that it can be trained to extract precedents and procedures from large amounts of unstructured text data. This means generative AI tools such as ChatGPT can be become a powerful assistant to lawyers, for example, in the scenario described below.
A lawyer receives instruction to structure a complex multi-jurisdictional deal and draft the relevant transactional documents. Unfortunately, they can only recall that they have done something similar but not when and for which client. In the normal course of events, they will then proceed to search their document management system or precedent database. This process can take an hour or more and may yield nothing if the lawyer cannot recall details of the past transaction.
Assuming the law firm has in place a precedent / knowledge system powered by generative AI, the lawyer merely needs to type in plain English a description of the deal or the legal issue and the tool can instantaneously identify the precedent and even others done by other partners. The tool can also assist lawyers to find out how other clients have been advised on similar legal issues by the firm, and find the firm partner / expert on a topic or do legal research.
Having access to such a KM tool is beneficial to lawyers and clients, in terms of time and cost management. It enables lawyers to spend their valuable time focusing on higher value (such as fine tuning the transaction documents) and more strategic guidance work rather than on drafting and research.
Legal Industry Adopting Generative AI on a Large Scale Soon? Unlikely
Just like other AI technology, it will face implementation challenges such as costs. Full data teams comprising data scientists and software engineers are required to support the development and use of generative AI, a high cost investment that many law firms are usually reluctant to bear.
The complex structure of generative AI also comes with a swath of ethical and security concerns. Lawyers who decide to use generative AI tools in their practice would be expected to understand how the chatbot works and the benefits and risks if they use it in their practice. They must be able to explain intelligibly the pros and cons to their clients.
Issues relating to confidentiality and conflicts of interests also come into play. Potential data bias related risk inherent in any AI tool is another concern.
AI is trained with human input and there are numerous examples of how bias has been introduced into algorithms. Drawing conclusions from AI may include implicit bias that could influence an argument, disadvantage a client or reinforce societal biases.
OpenAI’s ChatGPT has access to massive amounts of data sources to train its model effectively. It will be difficult and daunting for most law firms to replicate this training as they lack the hardware and internal IT infrastructure to manage the massive data quantity required to achieve comparable results.
ChatGPT Will Not Replace Lawyers
One of the main reasons why people still engage lawyers is because of the accountability and assurance that comes with a human lawyer’s advice on legal issues. It is relatively easy to obtain answers to most legal questions from ChatGPT. However, the quality of responses from ChatGPT can vary from accurate to misleading, albeit always delivered confidently. This is unacceptable as precision is critical in legal matters.
ChatGPT lacks authority. Its answers to legal questions usually do not include footnotes or specific references. It may mention a case name but no citation. Additionally, the AI is a “black box” as it does not explain the legal reasoning it took to reach a conclusion.
ChatGPT is not reliable. While experimenting with ChatGPT, I discovered that it could be manipulated to create false answers or the answers that you are looking for. ChatGPT has been shown to confidently provide answers that are completely wrong, known as “hallucinations” (made up facts). This is disconcerting because fluent and coherent answers lull users into a false sense of security, resulting in users not noticing errors or omissions because the response appears to make sense.
Even as the accuracy of ChatGPT improves, it is still a long way from providing value added commercial guidance that only human lawyers can provide.
ChatGPT currently lacks the nuance necessary to create consistently accurate responses and complex legal arguments. As such, it is not in a position to replace lawyers, not in the near future.
The Future of Knowledge Management
Only time will tell what role ChatGPT may or may not play for the legal profession. ChatGPT is still a prototype application. It will get better, and it will invite competition. There will be many other NLP models similar to ChatGPT.
More clients will use ChatGPT and they may come to a law firm and ask why the law firm has not incorporated generative AI technology into its practice. There are already law firms and legal service providers in the UK that have started using their own data or other proprietary data with the technology.1
In the coming months, ChatGPT will become a “starting point” for clients to educate themselves on legal topics. However, clients will still engage lawyers for specific legal advice, advice that requires soft skills, ones that ChatGPT does not yet possess. A positive outcome from this would be more efficient conversations between clients and their lawyers.
Given the increasing popularity of ChatGPT, it is critical that lawyers understand ChatGPT and be aware of its uses and limitations, so they can advise clients on when it is appropriate to use it and the risks of doing so. Hence, lawyers should not avoid or ignore ChatGPT. Instead lawyers should approach it with an open mind, embrace this technology and use it to their advantage, similar to how calculators help mathematicians.
The next five to 10 years will witness iterations of generative AI models with huge improvements, including addressing the current issues faced by ChatGPT. By the end of 2023, it is almost certain that many legal practices and service providers will announce that they have adopted generative AI technology in some form.
The adoption, development and implementation of generative AI in law firms will invariably involve knowledge management (KM) departments. The make-up of the traditional KM department in law firms will evolve. In addition to legal subject matter experts, data scientists, NLP experts and software engineers will form part of the KM team. There will be new roles required to improve and guide the generative AI system. An example would be experts who are able to engineer the appropriate prompts to guide the system in the right direction. A UK law firm has started recruiting for a GPT legal prompt engineer.2
The future of legal KM has never been more exciting.
Note: This article was written in the traditional way, including using pen and paper, with insights from the author’s personal experience and expertise on knowledge and risk management in law firms. Apart from experimenting with ChatGPT to understand its abilities, ChatGPT was not used to generate a first draft of this article.
Author: Angeline Poon
First published in the March 2023 issue of the Singapore Law Gazette