Opinion Menaka Guruswamy writes: AI chatbot, my future colleague

Let me illustrate by typing in questions to ChatGPT within my domain of the law. As a lawyer, I asked Chat GPT to answer a question in an area I practice — money laundering law. I typed the question: What is the offence of money laundering in India? ChatGPT not only identified the relevant statute — the Prevention of Money Laundering Act, 2002 — but also proceeded quickly to list the important ingredients of the offence, “proceeds of crime”, “knowledge”, “concealment”, “possession” and “projecting as legitimate”. It provided a few explanatory lines about each ingredient.

I thought I should ask something more complex and introduce a subjective element to the question. I asked, “Who is the best judge of the Supreme Court of India?” Here’s where the bot pleasantly surprised me. ChatGPT gave me a reasoned and institutionally appropriate response by saying, “It is not appropriate to designate a single judge as the ‘best’ judge of the Supreme Court of India. The Supreme Court is comprised of a bench of highly experienced and knowledgeable judges who work to uphold the Constitution and ensure justice for all citizens…The quality of their work is evaluated based on their judicial opinion, decisions and overall impact on the legal system. It is important to respect the impartiality and integrity of the judicial process and avoid making subjective judgements about individual judges.”

What I liked about ChatGPT’s answers was that not only were they informative, but also displayed an apparent commitment to being objective and refusing to pander to subjective biases. Significantly, it said that it is important not to make subjective or personalised judgments on individual adjudicators to protect the integrity of the judicial process, and so it is necessary to evaluate judges based on their judgments and overall impact on the legal system
.

However, it is necessary to vet ChatGPT-generated content. For instance, the Supreme Court does not comprise a “bench” (per its response to my question) but consists of many benches or panels of judges. This is unlike the situation, for instance, in the United States, where the court sits en banc — all the judges sit to hear each case. Significantly, one cannot regard ChatGPT’s responses as a final product to be relied upon, but instead should consider them solid first drafts. The responses must be looked over by an expert and corrected as required.

ChatGPT or similar AI tools are already being used by architects, lawyers and coders in their workplaces. For architects, it can generate images of designs based on inputs, for lawyers, it creates first drafts of basic agreements and can also swiftly write simple codes for coders. It’s also good with examinations. Recently, ChatGPT scored higher than many human candidates on an MBA exam at the very elite University of Pennsylvania’s Wharton School of Business. A week ago, ChatGPT passed parts of the US medical licensing exam and the University of Minnesota’s exams for some law courses.

It has so much commercial potential and is already so popular that on February 1, OpenAI announced the launch of a paid subscription version, called ChatGPT Plus, which will have a $20 monthly fee. The free version will continue, but the paid version will likely have faster response times from the chatbot and other enhanced features.

The Wall Street Journal reports that Microsoft Corp had announced that it will make a multiyear, multibillion-dollar investment in OpenAI, following up on earlier rounds of investments. Microsoft has said that it will incorporate AI tools like ChatGPT into all its products and make them available for other businesses to build on. The WSJ also reports that OpenAI has discussed selling existing shares in a tender form and will value the company at around $29 billion.

The future is bright for ChatGPT and other AI bots like it. However, while we will clearly embrace it for all that it offers by way of quick and succinct information, one must be careful about what it is allowed to replace. It is no substitute for thoughtful and careful research. Its findings cannot be relied on to substitute the judgment of a lawyer, doctor or engineer or other professionals. Finally, it might also start eliminating junior-level positions in fields like law, architecture, coding and research-oriented jobs.

Historically, the use of advanced machines in the manufacturing sector reduced the number of human workers needed and hence led to a consistent reduction of jobs in that sector. Similarly, the future of white-collar work will see a reduction in the number of humans necessary as AI will steadily replace them. And, herein lies the challenge for countries like India where unemployment is already high. How does one ethically deploy AI, while trying not to eliminate too many humans from the job market? Or is the inevitable future colleague a chatty bot?

Menaka Guruswamy is a Senior Advocate at the Supreme Court. She thanks the rather chatty ChatGPT for its responses that form part of this column