WHAT IS CHATGPT AND WHY SHOULD I CARE ABOUT ITS CAPABILITIES?
By Sean McCarthy, Co-Founder of ArbTech
Many people in the legal community will likely have come into contact with artificial intelligence (AI) and machine learning (ML) through tools used for eDiscovery, where these models search through the text of documents for keywords or patterns of language, and collate the results at an efficiency that a human could not match.
Some of us may also have interacted with chatbots when seeking online support for a product or service, but very few will have interacted with an AI language model like the one developed by the OpenAI team, which resembles the kind of communication portrayed in science fiction.
Interacting with ChatGPT gives an eerie sense that you are talking with a machine that does possess a lot of knowledge, but more importantly has a level of imagination and creativity that surpasses the scripted nature of any normal chatbot.
To get you up to speed quickly, ArbTech has outlined the possibilities and limitations ChatGPT could have for legal professionals in this article.
So what is ChatGPT? Let’s allow it to explain itself:
It is a variant of the GPT (Generative Pre-training Transformer) language model that is specifically designed for conversation and chatbot applications. It is trained on a large dataset of conversational exchanges and is able to generate human-like responses to a given input. One of the key features of ChatGPT is its ability to remember and incorporate context from previous exchanges in a conversation. This allows it to produce more coherent and relevant responses, rather than simply generating a response based on a single input.
ChatGPT Functionalities
When ChatGPT receives an input, it processes the text using its neural network architecture, which consists of a series of transformer blocks and a final linear layer. The transformer blocks allow the model to attend to different parts of the input and contextualize the response, while the linear layer is used to generate the final output.
In addition to generating responses, ChatGPT is also capable of performing various language tasks, such as translation and summarization. This makes it a versatile tool for a wide range of natural language processing applications.
Overall, ChatGPT is a powerful language model that is able to generate human-like responses to input in a conversational context, making it well-suited for use in chatbots and other conversational systems.
More importantly for the topic of this article, why should lawyers care about the capabilities of ChatGPT?
There are a few potential ways that lawyers could benefit from the capabilities of ChatGPT or similar language models:
1. Legal research and document analysis: ChatGPT and other language models could potentially be used to assist with legal research and document analysis by summarizing and extracting key points from large volumes of legal text.
2. Client intake and triage: A chatbot powered by ChatGPT or a similar model could be used to assist with initial client intake and triage by gathering information about the client's legal needs and directing them to the appropriate resources or legal professionals.
3. Legal document generation: Language models could potentially be used to generate simple legal documents such as contracts or wills, based on a set of input parameters. This could save time and reduce the risk of errors in the document preparation process.
4. Legal translation: ChatGPT and other language models could be used to accurately translate legal documents and other legal content, which could be useful for international law firms or for handling cases involving foreign languages.
Overall, the use of ChatGPT and other language models in the legal field has the potential to improve efficiency, reduce the risk of errors, and make legal services more accessible to a wider audience.
So, while ChatGPT is capable of extolling its own virtues, sceptics may say that these kind of responses are the least a language model that is supposedly as advanced as this should be able to produce. So what does it say about its limitations in use by legal professionals?
Wondering what practical implications ChatGPT might have for arbitration practitoners? Read about it here!
There are several limitations to the use of ChatGPT and other language models by legal professionals:
1. Dependence on training data: Like all machine learning models, ChatGPT is only as good as the data it was trained on. If the training data is biased or does not adequately represent the full range of legal issues and contexts, the model's output may not be reliable.
2. Lack of legal expertise: ChatGPT and other language models do not have a deep understanding of the law and cannot provide legal advice or representation. They can only generate responses based on the patterns and associations learned from their training data.
3. Ethical considerations: The use of ChatGPT and other language models in the legal field raises a number of ethical considerations, such as the potential for bias in the model's output and the risk of replacing human legal professionals with machine learning algorithms.
4. Limited to simple tasks: While ChatGPT and other language models are capable of performing a wide range of language tasks, they may not be suitable for more complex legal tasks that require a deep understanding of the law and legal principles.
Overall, while ChatGPT and other language models have the potential to assist legal professionals and improve efficiency, they should be used with caution and in conjunction with human oversight and expertise.
Let’s look at some more examples and try to get answers to questions more relevant to ArbTech:
Generate an arbitration clause with the legal seat being London, England.
"Any dispute arising out of or in connection with this contract, including any question regarding its existence, validity or termination, shall be referred to and finally resolved by arbitration under the LCIA Rules, which Rules are deemed to be incorporated by reference into this clause. The number of arbitrators shall be one. The seat, or legal place, of arbitration shall be London, England. The language to be used in the arbitral proceedings shall be English."
Generate an example of the dispute resolution provisions of a Bilateral Investment Treaty.
"Any dispute between the Parties arising out of or in connection with this Treaty, including any question regarding its existence, validity, or termination, that the Parties are unable to resolve through consultation or negotiation shall be referred to and finally resolved by arbitration in accordance with the rules of the International Centre for the Settlement of Investment Disputes (ICSID). The arbitration shall be conducted in English and shall be held in the capital city of Party A. The number of arbitrators shall be three, and the arbitration panel shall include at least one national of each Party. The award of the arbitration panel shall be final and binding on the Parties."
While Google could likely provide similarly high quality responses to the above questions, where ChatGPT impresses is through its ability to iterate and ‘learn’ from previous questions and answers, so that it can hone in on a more perfect response for the user. Let’s take the arbitration clause and ask the model to modify it further for our purposes:
Generate an arbitration clause with the legal seat being London, England, but instead of one arbitrator, the number should be three. Include in the clause a common form of nomination of the three arbitrators.
"Any dispute arising out of or in connection with this contract, including any question regarding its existence, validity or termination, shall be referred to and finally resolved by arbitration under the LCIA Rules, which Rules are deemed to be incorporated by reference into this clause. The number of arbitrators shall be three. The arbitrators shall be appointed as follows: Party A shall appoint one arbitrator, Party B shall appoint one arbitrator, and the two arbitrators so appointed shall appoint the third arbitrator. The seat, or legal place, of arbitration shall be London, England. The language to be used in the arbitral proceedings shall be English."
This flexibility and nuance certainly changes the game in terms of being able to supplement the workload and/or enhance the efficiency of a practitioner. The language model even has a form of imagination, whereby you can ask it to create or hypothesise on subjects that are forward-looking.
Design a dispute resolution system involving the buyers and sellers of digital assets through an online marketplace.
Here is a possible structure for a dispute resolution system involving buyers and sellers of digital assets through an online marketplace:
Self-Help Resources: The marketplace platform would provide various self-help resources such as a user-friendly FAQs, community forum, and a dedicated customer service team to help users understand the process of buying and selling digital assets, resolve any issues that may arise during the transaction and also provide a first-line support for any disputes that might come up.
Mediation: In the event that self-help resources are not sufficient to resolve a dispute, the marketplace would have a mandatory mediation process. This would involve a neutral third-party mediator who would work with the buyer and seller to try to find a mutually acceptable resolution. This process is meant to be fast, simple and cost-effective.
Arbitration: If the dispute cannot be resolved through mediation, the parties can opt to go for arbitration. The arbitration process will be administered by a reputable arbitration institution such as the American Arbitration Association or Singapore International Arbitration Centre. The parties will have to agree on the rules and the arbitrator(s).
Legal proceedings: If a dispute cannot be resolved through the above means, either party has the right to initiate legal proceedings through the courts.
Overall, the goal of this dispute resolution system would be to provide buyers and sellers with an efficient, cost-effective and easy to navigate means of resolving disputes related to the buying and selling of digital assets. The dispute resolution system is designed to be simple, fast, and accessible to all parties, with an aim to reduce the costs and time associated with disputes.
How will technology affect the future of justice in 100 words or less?
Technology is likely to have a significant impact on the future of justice. For example, the use of artificial intelligence and machine learning in legal research and document analysis could help lawyers and judges to more quickly and accurately identify relevant legal precedents and regulations. Additionally, the use of online dispute resolution platforms could make the legal process more accessible and efficient, particularly for small claims or disputes that do not require in-person hearings. However, there are also potential concerns about the potential for bias in the use of such technologies in the legal system, and it will be important to address these issues as they arise.
GPT-3 AI Chat: Imagination
But, the major caveat with ChatGPT in its current incarnation is precisely its ‘imagination’. The model will more often than not avoid giving a response in which it admits it does not know the answer, and instead will create what it believes should look correct to the untrained eye. And, due to the format and style of many of its answers, it is not easy to tell when the model is fabricating facts versus reciting them. The OpenAI team has released API’s to allow third party developers to build upon the foundation of ChatGPT, and we may soon see trustworthy web-enabled models, which can cite or link to the information used to provide responses.
Without doubt though, machine learning language models such as ChatGPT have now entered the public eye and built the foundation for their ubiquity in many industries, the legal industry being no exception. Drafting, document analysis and legal research are just some of the tasks that this technology will streamline, to the point where each practitioner may be able to increase their productivity by orders of magnitude.
In turn, the disruption of current legal business models may be just around the corner. The well-known startup DoNotPay has already begun the first steps of this disruption using ChatGPT, as noted in a recent article on The Swaddle. Their new service proposes to allow parties to wear earbuds that both listen to the oral arguments in the courtroom and supply supposedly accurate replies for the person to repeat to the judge and win their case. While there are a myriad of important ethical concerns around this specific type of solution, it becomes clear that is only a matter of time before we see machine learning-based language models like this being employed in ways we might not have thought possible today.
