OUR TAKE ON AI

The term ‘artificial intelligence’ (AI) was first suggested by John McCarthy in 1955, defining it as a challenge “of making a machine behave in ways that would be called intelligent if a human were so behaving”. Almost seventy years later, further to multiple waves advancing AI technologies and notwithstanding several so-called ‘AI winters’ (prolonged periods of time when interest and investment in AI was significantly decreasing), AI has finally arrived as an essential technology for our future development and is here to stay. Today, leading AI platforms are able to maintain logical conversations their users, thus, satisfying Mr McCarthy’s problem by making a machine behave intelligently. The benefits of AI for both individuals and businesses have transitioned from being purely theoretical to practicable and, to a great extent, quantifiable. For legal practitioners, presently, such quantifiable benefits would likely be based on the billable time saved, for example, on document review and textual analysis or production of documents based on standard templates. Further, there is a huge potential to use AI to write simple code automating mundane tasks, such as generation of exhibit lists, (re)numbering of exhibits, bulk-conversion of documents from one file format into another, updating cross-references or footnotes in a document — one can think of plenty of use cases and what is needed is a bit of knowledge on how to make basic changes to that code and run it. However, as of the date of this publication, it seems that the general consensus among legal practitioners is that AI systems cannot be reliably used for legal research and all of the results of such research would still have to be reviewed with great care by human lawyers. Galadari’s AI Commentary on arbitration rules, laws, and treaties, is an experiment focussed on using AI to ascertain the current quality of AI analysis, and to determine whether AI is able to digest large quantities of complex information and produce an accurate and logical analysis of the relevant text in respect of various arbitration rules. In the development of this AI Commentary, we used ChatGPT (GPT-3.5), an AI-powered language model developed by OpenAI. This AI model is capable of generating human-like text based on context and past conversations. The method used, for consistency, and not due to lack of imagination, implemented the use of the following prompt template to generate commentary on each provision of the relevant text: ‘analyse article/section X of arbitration law/rules/treaty’. Overall, the general outcome of our experiment indicates that, in the majority of situations, ChatGPT was able to successfully dissect provisions, spot important nuances, and generate meaningful commentary on the relevant provisions. Specifically, once ChatGPT was presented with a provision for analysis, it would typically attempt to break down the text into separate aspects into a list or bullet point format, tackling each and summarising the provision at the end of its analysis. This approach is consistently observed by ChatGPT across all the AI Commentary publications. We found this method to be suitable and informative, allowing readers to process the analysis in a systematic manner and benefit from ChatGPT’s separation of the provision into smaller bits that the reader may be specifically interested in. It was noted that ChatGPT does not overlook or miss certain aspects of a provision, which is a potential human risk factor. ChatGPT addressed every aspect of provisions without fail, for example:

Article 21.3 of the UAE Federal Law No. 6 of 2018 on Arbitration (“UAE Arbitration Law”): “The Arbitral Tribunal may, at the request of any party or on its own motion, amend, suspend, or cancel an interim measure it has ordered, in exceptional circumstances, by prior notice to be given to the Parties.”

ChatGPT: “Article 21.3 of the UAE Arbitration Law grants the Arbitral Tribunal the authority to modify, suspend, or revoke interim measures it has previously ordered during the arbitration process. These changes can be made under specific conditions referred to as “exceptional circumstances”. Let us analyse this article:

1. Modification: The Tribunal can amend or alter the scope or conditions of an interim measure based on the evolving circumstances of the case. 2. Suspension: The Tribunal can suspend the enforcement of an interim measure temporarily if the circumstances warrant it. 3. Cancellation: The Tribunal can completely cancel or revoke an interim measure previously ordered if it deems it no longer necessary or appropriate. 4. Exceptional Circumstances: The Tribunal’s power to amend, suspend, or cancel interim measures is limited to exceptional circumstances, implying that such measures are not subject to routine or arbitrary alterations. 5. Prior Notice: The article requires the Tribunal to provide advance notice to the parties before making any changes to the interim measures. This ensures transparency and allows parties to present their positions on the proposed amendments. […] The commentary produced using ChatGPT did display some shortcomings. For example, there were instances when ChatGPT would ‘hallucinate’, making something up, which is a well-known limitation not only for OpenAI and ChatGPT users but also for other AI systems, for example:

Extract from ChatGPT’s commentary on Article 2.4 of DIAC Arbitration Rules 2022 containing a reference to Article 32: “This exception indicates that certain specific provisions, such as Article 32, may have limited application to agreements to arbitrate made before the effective date of the Rules. It provides flexibility for parties to agree on different rules or procedures regarding the settlement or termination of proceedings, even if the general principle of the Rules taking precedence over the parties’ agreement still applies.”

Human lawyer: Article 32 of DIAC Arbitration Rules 2022 does not concern ‘settlement or termination of proceedings’ but sets out provisions pertaining to expedited proceedings.

Occasionally, ChatGPT would not appreciate the important distinction between ‘may’ or ‘should’ and the commanding ‘shall’, which can be crucial, for example:

Article 4.1(c) of DIAC Arbitration Rules 2022: “Any party wishing to commence an arbitration under the Rules shall submit to the Centre a Request which shall include: […] (c) a copy of any relevant agreement between the parties, which contains the agreement to arbitrate, in respect of which the claim arises; […].”

ChatGPT: “Article 4.1 of the DIAC Arbitration Rules outlines the requirements for submitting a Request to commence an arbitration under the Rules. Let us analyse each component of this article: […] Copy of relevant agreement: The Request should include a copy of any relevant agreement between the parties that contains the agreement to arbitrate, which is the basis for the claim being made.”

Parenthetically, ChatGPT sometimes would simply paraphrase a provision several times within a single answer or alter the format of the responses with a series of prompts within the same commentary. To be fair, most of these provisions were quite nominal and did not contain much substance for analysis, for example:

Definition of ‘Parties’ Article 1.1 of UAE Arbitration Law: “Parties: The Claimant and the Respondent, whatever their number.”

ChatGPT: “Parties: Refers to the Claimant and the Respondent involved in the arbitration. The parties could be individuals, companies, or any legal entities engaged in a dispute.”

It was expected that there would be such limitations, but it was pleasantly surprising that their frequency was much lower than anticipated. Being cognisant of all the existing issues and scepticism within the legal community, we consider that this experimental AI Commentary could be of assistance to practitioners in at least two ways. First, it allows the reader an opportunity to perform a sense check on their understanding of a provision or its aspect against ChatGPT’s analysis. Second, the time required to produce 24 publications comprising the AI Commentary was significantly less than the typical duration needed to produce a single comprehensive commentary text on any of the relevant arbitration laws, rules, or treaties. Thus, should it become necessary, a similar AI commentary could be produced on any arbitration law/rules/treaty at a fraction of time and cost typically associated with such a task. The purpose of publishing the AI Commentary is to provide arbitration practitioners and academics with a general sense of what is presently possible to achieve in the field of arbitration with the assistance of generative AI software and encourage the arbitration community to push the boundaries of arbitration as a flexible, efficient, and effective dispute resolution method. Notably, all commentary was generated with ChatGPT and was supported by a selective review by the Editors. Accordingly, the commentary may contain inaccurate and/or incomplete information. Readers are strongly advised to exercise caution reading the commentary with some scepticism and to keep a pencil in hand to note any inaccuracies. Needless to say, nothing in this text should be considered and/or relied upon as legal advice. For detailed information, please refer to OpenAI’s Terms & Policies: openai.com/policies. This project would not be complete without front page illustrations, which were also generated by AI. DALL E, another OpenAI system capable of creating images based on prompts, was used for this purpose. The chosen concept is based on a watercolour painting style, primarily portraying athletic rivalries in locations that correspond to the relevant arbitration law, rules, or treaty. The hope is that the readers will find the illustrations aesthetically appealing.