When OpenAI announced that it was releasing GPT-3 in June 2020, anyone who has been following the development of AI and natural language processing (NLP) was at once in a state of shock and excitement. Previously heralded too dangerous to release to the public, GPT-3 has been described as game-changing by numerous industry observers, and the technology stands to make an impact in the legal profession just as it will in other industries like sports journalism, web development, and social media marketing.
What, though, is the reality of that ‘impact’ likely to be? Are lawyers going to be put out of business by a machine? Or is this just a distortion of what otherwise could be a very useful tool for legal practice?
What is GPT-3?
GPT-3 is a neural-network-powered language model. In plain language, it is technology that is able to ‘learn’ about language (from intensive human programming based on 175 billion parameters) and then make highly accurate and context-specific predictions about what would come next in a sentence given any particular prompt.
In other words, GPT-3 is the closest thing we have to a machine writing like a human.
While there are many excellent explainers elsewhere about what GPT-3 is and how it works, the bottom line for lawyers is that now a technology exists that not only can learn new material significantly faster than any human, but that can emulate writing styles so accurately that many humans cannot differentiate between material written by humans and material written by GPT-3.
How will GPT-3 impact the legal profession?
Speculation around how GPT-3 will impact the legal profession has been intense, and indeed somewhat revelatory of how stable (or not) lawyers see their own careers to be in future. While many legal technologists tout the benefits of innovations like artificial intelligence in legal practice, the remainder see this as a threat to their livelihoods.
In the examples that follow, I take a look at two realistic applications of GPT-3 in legal practice: legal research, and legal information-sharing.
Somewhat boldly, AI optimists predict that GPT-3 will not only impact the way that the lawyers do legal research, but that it will democratize legal research.
This claim, however, seems to be based on a statement made by Andrew Arruda, CEO of ROSS, an AI-powered legal research platform. He said: “Thanks to GPT-3, you don’t have to be a machine learning scientist with millions of dollars to access an extraordinarily powerful natural language model. You just need to be an innovative tinkerer with access to OpenAI’s API.”
The way I see it, legal research is not being democratized by there being more accessible means of harnessing the power of NLP; it’s just that legal research platforms will, in the coming years, become so much more advanced that the speed at which we do legal research will be improved exponentially.
LexisNexis have been building intelligent Q&A capabilities for years, and it stands to reason that in future we may be able to input questions into a platform and get highly relevant and semantically accurate responses that effectively answer legal questions. How we harness these in serving clients, though, is another matter entirely.
It is a fundamental tenet of access to justice that anyone wanting information on the laws which he or she is expected to obey should be able to access it, yet the profession’s monopoly on the commercialization of legal knowledge has of course continued unabated. GPT-3 might be set to change this.
At the present time, we are far off having a machine replica of a lawyer that would be able to share information about the law upon request… but we may not be as far off as we think.
Chatbots represent our closest current achievement to that eventuality. With an appropriately programmed chatbot, a user would be able to ask for advice on a legal issue, using their own language, and the chatbot would then be able to translate that natural language into legal language such that it could perform detailed searches in legal databases, ascertain the appropriate outcome for their problem, and then relay that advice to the user in their own (natural) language. If this were achieved fully, this would be truly revolutionary. The reality, however, is much different.
Even though we are getting much closer to being able to have a chatbot that could readily respond to clients (say), we are far off being able to use this technology ethically. There will, of course, be mistakes in the analysis made by the AI, and accordingly there will be inaccuracies conveyed to end users. When it comes to the remedy of legal problems, this is intolerable. For this reason alone, we may be many years from having a world where anyone could talk to a computer to resolve legal problems.
Where next for GPT-3 in the legal profession?
GPT-3 is still in private beta and the range of applications for which it could be used have not yet been fully fleshed out. This includes in the legal profession. With relatively few people having access to this technology, it is safe to say that we do not quite know what GPT-3’s impact will be. It certainly will not be the end of lawyers, but it might be the end of lawyering as we know it.