Benjamin Alarie (Toronto; Academic google; Managing Director, Blue J Legal), The Rise of the Robotic Tax Analyst178 Fed. Tax Notes 57 (January 2, 2023):
As a bold tax writing experiment, this installment of Blue J Predicts has been generated with the help of an artificial intelligence assistant, OpenAI’s “Generative Pretrained Transformer 3” (GPT-3). GPT-3 is a large language model developed by OpenAI and supported by Microsoft. He is an inexhaustible text generator and can write accurately in English on almost any topic. He has fulfilled his duty, with my human company, tirelessly.
This isn’t the first time a legal scholar has invoked a robotic co-author, and I expect these kinds of tools to become more and more common. I hope they will eventually be as noticeable as using a spelling or grammar checker. However, at this time, before the rise of the robotic tax analyst, the use of GPT-3 to help write this article is likely to come as a surprise to some.
In October 2021, I produced a peer-reviewed law review article with my academic colleague, the late (and great) Arthur Cockfield Professor of Tax Law at queen’s college. The article was notable for being the first peer-reviewed law review article to take full advantage of GPT-3 in its production. In that article, after a brief introduction written by us, Benjamin Alara and Cockfield: We gave GPT-3 control of the metaphorical keyboard and allowed it to produce its textual analysis without interruption or editing.
The results were mixed and intriguing, and certainly pointed in the direction of a future possibility. In our opinion, GPT-3 had potential. In the article, we speculate on the future of AI in legal scholarship, provocatively asking in the title, “Will machines replace us?” Cockfield and I conclude that “although GPT-3 is not up to the task of replacing law reviewers today, we are much less sure that GPT-5 or GPT-100 are not up to the job in the future”.
Although these words were published just over a year ago, technological developments in 2022 suggest that we were probably somewhat conservative in our measurement of the likely rate of AI improvement in the form of large language models. The rapid progress in power of these models has been quite impressive, with more progress to come. To help show the power of GPT-3 to readers of tax notesI think I am the first author of these pages to have the help of GPT-3 to write a contribution.
As 2023 begins, developments with large language models, most notably ChatGPT, a recent offshoot of GPT-3, are generating excitement in legaltech circles. They are also raising caution among some in academia and legal practice. On November 30, 2022, OpenAI made ChatGPT available to the public through an open beta version. In less than a week, the system was implemented by more than 1 million registered users. The extremely fast adoption was fueled by claims, and plenty of evidence, of impressive algorithmic feats performed by ChatGPT. ChatGPT’s capabilities reportedly include achieving passing scores on practice bar and medical board exams.
Within days of ChatGPT’s launch, Andrew Perlman, dean of law and professor at Suffolk University, urged the program to write an essay about its own likely influence on the future of law. [The Implications Of OpenAI’s Assistant For Legal Services And Society]. The abstract of the essay, which Perlman claims he wrote himself, concludes with the observation that “interruptions in the rapid development of AI are no longer in the distant future. They have arrived, and this document offers a small taste of what is to come”.
It’s not just GPT-3 and its offspring that have been making waves. Other algorithms garnered media attention in 2022 for exhibiting surprisingly strong imaging capabilities. These latent text-to-image diffusion models, including DALL-E 2, Stable Diffusion, and Midjourney, have been widely used, critiqued, implemented, and celebrated.
The New York Times reported in September 2022 that an image generated by a text-to-image broadcast model won a digital art competition at the Colorado State Fair, causing a stir among artists, competitors, and social media users. It’s fair to say that machine learning and AI are starting to make their influence felt in some unexpected places. There is little reason to expect these and related developments to abate on the threshold of fiscal analysis.
In conclusion blue j Predicts for 2022, we will do a Dickensian balance, visiting blue j past, present and future predictions. The initial agenda is to take a look back at how Blue J Predicts analytics have fared in 2022. Have our predictions come true? Have the cases gone the way the machine learning models and algorithms have anticipated? In general terms, the answer is yes, although with the important caveat that we still do not have the last word of the courts in many of the cases.
We then consider the state of AI and machine learning in tax research and analysis. We outline some new ways in which machine learning technology is poised to impact tax research and analysis this year, in blue j and beyond. Finally, we look to the future to anticipate how AI could affect tax analysis and research, and explain why it’s important to start focusing our attention on the possibilities.
Previous TaxProf blog coverage:
https://taxprof.typepad.com/taxprof_blog/2023/01/alarie-the-rise-of-the-robotic-tax-analyst.html
