Monthly Archives: March 2023

Mishcon de Reya GPT Legal Prompt Engineer

It’s extraordinary how quickly businesses are starting to adopt GPT technology to get a competitive edge. Professional services such as lawyers, accountants, consultants, actuaries will need to move fast if they are not going to be left behind. Much of the grunt work done by these firms, work that is typically charged out at pretty high fee rates and done by entry level staff, can now be done faster by AI tools and, as technology advances and improves exponentially, better and a great deal cheaper. In London, Mishcon de Reya have recently advertised for a Legal Prompt Engineer which is interesting not least in terms of how rapidly they’ve moved.

How then are the current fee levels going to hold up? Will the introduction of AI tools such as ChatGPT and now Bing with Chat finally start to deliver lower fees? Professional services have been notoriously resistant to downward pressure on fees but if a good legal prompt, especially when done in conjunction with an upload of key case documents, can deliver the sort of analysis in seconds traditionally done by junior lawyers in days, there’s some hope for buyers of legal services. It won’t stop senior lawyers charging a fortune but it may reduce billings from junior staff.

At vCAIO we can’t see professional services using the existing pyramidal structures of seniors supported by a broad base of lower tier professionals much longer because it won’t only be the lawyers who are using AI – their clients will be using AI too which means a lot of what they are currently paying for will be available to them directly from their own desktops. Once businesses master the use of prompts they’ll start to ask their professional advisers how much they rely on prompts to deliver the services they are paying for. This ultimately could lead to a reduction in the number of low and mid tier professionals whilst potentially enhancing fee rates for the most able senior partners. The days of charging huge fees for hours of work by junior staff could be coming to an end. Afterall, what is the hourly rate for ChatGPT?

It’s currently 19 cents and is available 24/7.

Chief AI Officers will become key C-Suite executives

Virtual Chief AI Officers will be key C-Suite members in successful businesses for 3 reasons.

AI will be exploited in all business departments from marketing to production, from accounting to legal, from HR to strategy, from capital allocation decisions to mergers and acquisitions.
Business will need to do things better, faster and cheaper to survive and remain competitive.
Those that exploit AI most effectively will be the winners.

Utilising AI tools to maximum effect will demand Board level representation to make sure its adoption is sponsored across the organisation.

Will there be a conflict with IT? Yes. But smart CTOs will already have recognised the importance and value of AI as a code-writing accelerator, as a security enhancer and as client support agent that will reduce the demands made on expensive staff whilst freeing them up to focus on higher value activities.

Crafting effective ChatGPT prompts

As an AI language model, ChatGPT is designed to understand and respond to human language in a way that is as natural and intuitive as possible. However, the accuracy and relevance of the responses generated by ChatGPT depend largely on the quality and specificity of the ChatGPT prompt provided by the user. In other words, the more detailed and specific a prompt is, the better ChatGPT can understand the user’s needs and provide more relevant and helpful respon§ses.

One of the main reasons why specificity is so important is that natural language is inherently ambiguous and context-dependent.

For example, a simple prompt like “What is the weather like?” can be interpreted in many different ways depending on the user’s location, the time of day, the season, and other contextual factors. If the prompt is too vague or generic, ChatGPT may struggle to provide a response that is accurate or relevant to the user’s specific needs.

On the other hand, if the prompt is more specific and detailed, ChatGPT can leverage its deep learning algorithms and vast knowledge base to generate responses that are tailored to the user’s unique circumstances. For example, a prompt like “What is the weather like in New York City right now?” provides ChatGPT with a much clearer context and allows it to generate a response that is more accurate and relevant to the user’s needs.

Another reason why specificity is so important is that it helps to reduce ambiguity and confusion in the communication process. When prompts are too general or vague, ChatGPT may generate responses that are not entirely relevant to the user’s needs, leading to frustration and confusion. By providing more specific and detailed prompts, users can ensure that ChatGPT is better equipped to understand their needs and provide responses that are more accurate and helpful.

Moreover, the specificity of the prompt is crucial when it comes to complex or technical topics. For instance, if a user is seeking medical advice, a prompt like “What should I do if I have a headache?” may not be sufficient to provide them with the information they need. However, a prompt like “What are the possible causes and treatments for migraines?” can provide ChatGPT with a much clearer context and allow it to generate more detailed and accurate responses.

Prompt Specificity

The specificity and detail of prompts are essential for the effectiveness of ChatGPT. The more detailed and specific the prompt is, the more accurate and relevant the responses generated by ChatGPT will be. By taking the time to provide specific and detailed prompts, users can ensure that ChatGPT understands their needs and provides them with the most helpful responses possible.

Incidentally, the fact that ChatGPT can understand highly specific English language prompts is often overlooked as a major leap forward in AI. The internet is abuzz with arguments about the quality of ChatGPT’s output but what in many ways is just as remarkable is how well it deals with inputs. ChatGPT’s ability to comprehend complex, highly specific prompts is a totally new phenomenon. We will be chatting with ChatGPT very shortly – beats the hell out of Siri.

Prior to AI tools like ChatGPT it used to be the case with computers that “garbage in, garbage out”. In other words, if you put in bad data or write bad programs or construct bad searches, you get bad outputs. Consequently you had to craft your instructions with a lot of thought and care. As this article about Chat GPT and GIGO explains, this no longer applies.