In this interview, Michael Perkins and Jeremy Duffy of Autonomy First Lawyers give us their take on ChatGPT, how it is impacting the legal world and how it can help our clients.
Q: Can you outline for us in simple terms what Chat GPT is?
Michael P: ChatGPT is a very advanced chatbot. It is an app that you can ask a question and then ChatGPT and the platform that sits behind it will answer back and give you a result. It uses a technology called generative artificial intelligence, and the significance of this is that it’s actually pretty darn good at what it does when you are asking general questions. It’s creating a lot of interest because it does give good responses. It’s not perfect, but we can draw a parallel with the start of the Internet. The commercial Internet, as we know it today, is built on the foundations of academic research with Netscape as the first widely-used browser. That’s about where I think we are with AI at the moment and ChatGPT represents the Netscape of the generative artificial intelligence world.
Q: How is ChatGPT useful to us? Why should we care?
Jeremy D: We should care in two ways. Firstly, in terms of how ChatGPT may be useful in our everyday lives, in the legal area for example, it can give us the ability to provide legal information and complete other legal tasks in a very efficient manner that improves the access to information and legal services for our clients. This could also work in medicine, finance, and all areas of work and play.
We should also care, however, because there are some important social issues to consider around how we use AI, such as ChatGPT. There’s obviously a concern about how many jobs it might take, and how those jobs might be reconfigured. In the academic world, we need to manage what students are learning because artificial intelligence is writing answers to exam questions and things like that.
So, there are two reasons to care. One is on the positive side, but there’s certainly a negative side that needs to be taken into account.
Q: How should we be using ChatGPT? Is it simply a research tool?
Michael P: The simple answer is no, but to expand on that, we need to look at ChatGPT as a system whose commercial life is currently being proven out. There’s an area of function with ChatGPT that is immediately useful. However, there are also areas where it’s concerning, because it’s not yet good enough, and that’s where the academic world has its concerns. I think the question we need to ask, as this tool is still coming out of the labs, is, for what is it most fit for purpose? And I think that the simple answer to that is it’s best just as a source of information, not as a source of truth. This makes it different from Google, which will give you an index of scholarly articles, that are reasonably accurate. ChatGPT is being used in marketing and as a tool for creating creative content and that’s what is getting the schoolteachers worried.
The best answer to the problem that I heard from a teacher was, ‘Oh I’m not worried about ChatGPT because the way we are going to use it is to tell the students to use ChatGPT and then give a critical response as to why that answer is good or bad’.
I think this is a good reminder not to be reflexively critical of the tool. It’s like the old saying goes, guns don’t kill people, people kill people. So let’s not get too up in arms about the tool itself, let’s focus on the most valuable ways it can be used.
In practice, we are advocating that our clients use ChatGPT if they want to learn about general problems or ideas that they have, and then come to us and talk about the results, their concerns and how we can help.
The applications are endless. A couple of weeks ago faced with having to make a speech at my daughter’s engagement. I asked ChatGPT to tell me what to say about my daughter’s engagement when she’s had a romance for nine years and she’s been living with this fellow for 18 months. The result was the most pleasing prose that I could have imagined, a speech that I couldn’t have written better myself. So that’s a good example in the social context of how ChatGPT can help clarify your thoughts, improve your expression and get your message over really well.
Q: What are the biases inherent in ChatGPT?
Jeremy D: ChatGPT is only as good as the data that it has seen. That is the same for all AI products, which generally use statistical analysis, working out what’s the most likely next word in a sentence based on the nature of the question and the information that they might be looking at in the background. And so, if the background material has biases, then ChatGPT and other AI products will have biases too, and this may not be taken into account by the user. AI needs to be looked at as potentially biased and it also needs to be looked at as potentially rubbish because it’s not using any critical analysis of the material that it’s presenting.
Another problem is, it can pick up on the bigotry that’s available on the internet and use this information to create abusive responses. This was especially a problem in the early stages of ChatGPT. It doesn’t always have the capacity to determine which parts of the information it uses are good or bad or indifferent. It just statistically analyses that information and says this is probably what they want me to say. It has been trained to some degree to weed out that sort of information, but it’s still there to an extent. You should always take the answers with a grain of salt until proven to be true.
Q Is ChatGPT dangerous?
Michael P: We can’t personify ChatGPT and give it human characteristics, saying that it is malevolent. It doesn’t have the human intelligence to be malevolent.
You may as well ask is a rifle or a handgun dangerous? I think the answer to that simply is yes. If it’s used inappropriately. But billions of dollars are being invested in this technology, so we can see that it is not going away. We need to run towards it rather than run away from it, but we must also take responsibility for establishing rules and parameters for the appropriate use of the tool.
This comes back to families, businesses and professionals looking at ChatGPT and asking, how do we make sure that we use this in a way that adds value to ourselves and clients or customers? We have to provide the critical appreciation that Jeremy was just referring to.
Step one to create this safer environment is keeping the human at the centre of the decision about how ChatGPT is to be used. We need to keep reminding people that ChatGPT is only a tool. We have to provide the purpose and the context for a safe operation. That’s where we really sit as technology enabled and supported lawyers. We are the ones who need to work out how can we use this in a smarter way to add value to our services and our clients.
Q How is ChatTGPT useful in the legal world?
Jeremy D: ChatGPT and other AI-based tools will, in my view, revolutionise how lawyers work and how legal services are delivered to law firm clients. The technology has a way to go before this potential is fully and safely realised, but the path that this technology is on is undeniable. The overall effect of ChatGPT on the legal profession will be the automation of base-level legal services, including the provision of generic advice and the automation of legal output, such as documents, pleadings, and the like.
To expand upon that, ChatGPT is likely in the future to be able to provide highly accurate, generic advice without lawyer intervention. As I said, it’s not quite there yet, but with specific training on massive volumes of legal information, I’ve no doubt that ChatGPT, or whatever AI programme it might be, will get there in time and probably pretty soon, in years rather than decades.
To me, it’s a bit like the driverless car. Everybody thought it was going to come in very quickly, but then all these legal issues had to be resolved around it. Once those are resolved, though, things will change quickly. Of course, lawyers will always be required, to refine and tailor any advice that’s provided to the client through ChatGPT in order to take into account the client’s specific circumstances, and provide the nuanced and strategic advice that lawyers should be providing to their clients. However, AI such as ChatGPT will become an indispensable part of the overall process of providing that advice.
An example of the area of automated legal output is contract creation and execution, which is an area that I’m particularly interested in. ChatGPT is a huge step forward in the provision of what are called smart or self-executing contracts. ChatGPT has already demonstrated the capacity to create written contracts from a narrative style data. It can also then write the code that is required to automate the transactional aspects of that contract.
Lawyers everywhere will soon be creating self-executing everyday contracts easily and efficiently and incorporating this into contract lifecycle management software. That’s not to say that there won’t be issues to be resolved in this regard, there will be many issues, but if the output is worth that, then the way will be found to do it.
Q: So how can a client of the firm make good use of ChatGPT today?
Michael P: ChatGPT can give people the opportunity to interrogate and think through a problem more deeply before talking to us. So, I think the best use for clients today of ChatGPT is to ask the question that they’re most concerned about. Then when you get to the point where you feel you understand the problem better and you want to talk to us, please come in. It takes us beyond the Information Sheet, with static information, and gives people a tool to dynamically explore a problem.
I would encourage clients, if they want to go down this route, to go and open an account on the open AI site. The account is free and it’s a web-based service. You ask the question. It’ll give you a response and you can see what you think. And then, with a greater understanding of the issue, you can come to us and we can talk about the question, the response and what we can do to help you with this problem. I think that this can help us get to the bottom of a problem and allow us to help clients more quickly and cost-effectively. Let’s not be afraid of ChatGPT, let’s give it a go. But the lesson in all of this is to keep humans in the decision-making cycle. Let’s make sure that we use ChatGPT as a tool to refine our enquiry and our thinking, but not as a source of truth.
To start using ChatGPT go to https://openai.com/blog/chatgpt/ and follow the prompts.