After ensuring every participant had access to GPT, the workshop commenced with Kay posing a rather obscure question. “If I’m a VC and you’re an entrepreneur with a time machine, what year would you choose to pitch a pocket calculator business to me?” she asked. As you might expect, the group drew a blank, but this is where ChatGPT began to flex its muscles. By simply typing Kay’s question in and hitting ‘enter’ the group had a solid answer in a matter of seconds.
“The early 1970s”, suggested one of our members. “Great”, Kay replied, “but why?” This time, the group immediately started typing, and in about 5 minutes they had a nuanced, compelling pitch – based on technological developments and cultural shifts – that identified 1972 as the best year to start a pocket calculator business. Quite an impressive feat for a group of people who knew next to nothing about the pocket calculator and its history just 10 minutes prior.
Within the first five minutes of the session, Kay demonstrated just how powerful ChatGPT can be in aiding our understanding with unparalleled historical and cultural knowledge. It was the practical nature of this task, however, that had the biggest effect. “I avoided GPT because it sounded so complicated”, admitted one member. “I thought it would take too much time to learn how to use it, but now I’ve seen first hand how easy it is, it’s definitely something I’m going to continue with.”
The (Sort Of) Unparalleled Abilities of LLMs
Another common misconception when it comes to generative AI is that human intelligence can no longer compete. This is a tool that can sift through the entire contents of the internet in mere seconds, so anyone who has that misconception can be forgiven, but the truth is that generative AI is far from perfect. Understanding what it’s good at, and what it isn’t, is a crucial part of maximising its potential, which Kay explained to the group.
“Think about the work GPT has done for you in that task”, Kay instructed. “What has it made easier?” The first answer from the group referenced the historical and cultural knowledge GPT displayed, and this is where GPT really excels, according to Kay. “The historical knowledge GPT has access to is unparalleled in any one person”, she clarified, “and in that sense it is certainly superhuman. It also processed your intent, analysed the data it had access to, and made a judgment as to which year would be the best option based on that analysis.”
Kay continued to explain that LLM tools like Chat GPT should therefore be thought of as a ‘calculator for thinking’, a tool that can speed up certain tasks and offer answers that we would find difficult to come to ourselves. “But that only applies to certain types of thinking”, she revealed. “It’s great at analysing, reasoning about, or processing human culture, but it is prone to hallucination, slow to update its knowledge, and is relatively slow to use.”
These limitations mean that users need to be selective in their operation of ChatGPT and other LLMs. You wouldn’t reach for a calculator to work out how much change you’re due at the newsagent because you can do the sums quicker in your head. You would, however, reach for it when you’re completing your tax return or updating your P&L sheet, and the same applies to AI.
When to Reach Out for GPT’s Help
So, how do you know when GPT will help and when it will hinder, especially if you have limited experience with it? “I like to think of it as having a box filled with 100 miniature people that you can ask questions of or instruct to complete tasks”, Kay explained. “If I’m dealing with something that 100 people would be able to do better or quicker than I can, then reaching out to them isn’t just helpful, it’s strategic.”
Whether it’s gathering historical data, exploring various perspectives on a subject, or providing a template for a document, for instance, GPT is likely to save you a great deal of time. If, on the other hand, the task is simple and straightforward, or deeply complex or personal, the unique human touch and intuition are irreplaceable.
In scenarios where you’re confident in your own knowledge or have access to an expert, or where a human connection is key, stepping away from GPT is advisable. After all, while the technology is advanced, it doesn’t possess the nuanced understanding, empathy, and personal experience that human intelligence brings to the table.
In essence, GPT should be viewed as an extension of your capabilities, not a replacement. It’s a tool designed to enhance your workflow, but it’s also crucial to recognise its limitations and understand that the human element you provide is invaluable. Knowing when to open the box and when to rely on your own expertise is an art in itself, one that is best refined through personal experience.
Three Things We Learned
Kay’s workshop was a real eye-opener for our members, proving not only the effectiveness of GPT, but also the ease with which it can be harnessed. To distil the group’s learnings, here are three things – three pieces of actionable advice – you can use to get started with ChatGPT: