EY refers to the global organization, and may refer to one or more, of the member firms of Ernst & Young Global Limited, each of which is a separate legal entity. Ernst & Young Global Limited, a UK company limited by guarantee, does not provide services to clients.
Recent Searches
Generative AI tools such as ChatGPT can help provide previously impossible services / products, thereby enabling businesses to deliver long-term value.
In brief
- ChatGPT is not limited to a specific mission and is a more general AI tool than a purpose build chatbot.
- The tool can usher in a new era in hyper-personalisation, especially in e-commerce and consumer product industries.
- ChatGPT is not meant to be a tool for automating conversations. It rather uses conversations to automate tasks through conversations.
Much has been written about the power of ChatGPT and the potential risks it presents in a variety of fields including education and financial fraud. However, its potential benefits are even greater if its power is harnessed properly and if organisations consider using low-risk or zero risk applications.
To begin with, it is necessary to understand what ChatGPT is and the uses to which it is best suited.
ChatGPT is an AI-based tool which can both understand and generate natural language. This means that we give it queries in normal speech and get back a very human-like response.
ChatGPT has been trained using vast amounts of data, which is why such tools are called Large Language Models (LLMs). This means that ChatGPT also has vast reserves of knowledge. Hence, we can ask very specific questions such as “how would you describe aerodynamics to a five-year-old” and very often get answers which are extremely relevant.
A versatile chatbot
It is, therefore, quite natural to ask if ChatGPT is just a clever chatbot. While it is a chatbot, the ones we typically interact with tend to be built and trained to serve particular purposes, in areas such as customer service, and answer a narrow range of quite specific questions. ChatGPT by contrast is not limited to a specific mission and can attempt to answer almost any question put to it unlike other chatbots, making it a more general AI rather than a replacement of the current conversational automation technologies.
One example of a use case is hyper-personalisation in e-commerce. ChatGPT can be asked to take descriptions of different products on a website and re-write them to appeal to different age groups and market segments. This opens up the possibility of websites being able to present customised text for individual visitors by detecting their demographic and other personal information and calling up the descriptions created by ChatGPT. The same concept can be applied by consumer product companies as well.
ChatGPT can also be asked to develop software codes. For example, it could be asked to create a piece of code in the C++ programming language which would remind a computer user to take a break at specified time intervals. This is well within the capabilities of ChatGPT and while it is not always successful in writing working code for more complex scenarios, ChatGPT can do the initial code writing work, leaving it to the developer to add any polish required.
Hyper-personalisation - a key differentiator
The potential use of ChatGPT to replace humans by delegating tasks previously carried out by them has been much discussed but is not the optimal use of the technology. It is often referred to as the AI fallacy. Generative AI tools such as the ChapGPT never tire and can access almost instantaneously all of the vast reserves of knowledge at their disposal. Why limit it by replicating what we do as humans when it can do things that we wouldn’t consider asking a human to do?
The hyper-personalisation use case is a very good example. In that instance, it would not be necessary to pre-load the different product descriptions into the website’s memory. ChatGPT could simply develop them as and when a particular user looked at them, meaning that no two descriptions of the same product will ever be quite the same.
It is also important to understand that ChatGPT is not a tool for
automating conversations. Rather, it automates using
conversations.
This is a capability that lends itself to the transformation of products and business models, enabling organisations to rethink how they work. ChatGPT can automate tasks which we couldn’t do even with humans and, hence, the tool enables organisations to deliver new services and products to clients in new ways.
What are the risks?
The risks associated with ChatGPT and other AI technologies have been the subject of much media attention in the recent past. However, it is important to acknowledge that the risks lie in the use to which it is put, not in the technology itself.
Generally speaking, the risks arise when the technology is used for a purpose to which it is not inherently suited, such as writing a newspaper article unaided. This is not a good use case for ChatGPT for several reasons, the most important one being it may not be possible to audit and verify the source from which ChatGPT curates its information.
It can also exhibit all the traits of hallucination where it believes something to be real even though it is not. This can arise because it merely generates a textual response which it finds relevant to a query, rather than drawing information from a verified set of data. Neither does ChatGPT possess the healthy human scepticism which could sense check it on an instinctive basis. More importantly, ChatGPT is not creative in itself. It relays the ideas and knowledge in the data on which it has been changed.
Social media and the internet contain vast amounts of false information which, without the aid of human fact checking, a machine can consider to be true. Therefore, any use of ChatGPT for news writing would have to be overlaid by rigorous human fact checking.
Delivers business benefits
Having said that, there are some clear areas where Irish businesses can benefit from ChatGPT.
ChatGPT is just one of the tools that is democratising AI, allowing business users with no background in IT to utilise the power of AI in their day-to-day work. We already see employees across many different industries and business functions taking the initiative to start using ChatGPT to make their jobs easier.
ChatGPT has the potential to deliver a range of business benefits – from making improvements to call centre operations to elevating product or service designs.
A significant challenge for call centres is that a lot of their data is unstructured. For example, recorded calls are not directly useful as data unless a lot of painstaking work is put into reading and understanding them. ChatGPT can help to summarise calls or even just extract key issues presented by a customer. This allows call centres to structure enormous amounts of unstructured data and transform themselves from cost centres to being a vital source of data for the businesses.
A very valuable use case lies in product or service design.
ChatGPT can be used to harvest user sentiment data in relation to
products or services and suggest tweaks or improvements to make
them more appealing to certain customer cohorts.
The technology can also be extremely useful for product descriptions and documentation. For example, when preparing the content for product booklets, manuals, container labels, and so on, ChatGPT will be able to help improve quality and time to market by generating first drafts at high speed as well as re-writing them for specific purposes and markets.
What all these use cases have in common is that they are not automating conversations, which is where many organisations assume that it can create impact. These are use cases where conversation can be used to automate specific tasks, using the capability of ChatGPT to understand natural conversational queries and respond in the way we want it to. The only limitation, it would appear, is the imagination of the humans using it.
Summary
ChatGPT can exhibit all traits of hallucination, but it is important to acknowledge that the risks lie in the use to which it is put, not the technology itself
Related articles
How organisations can adopt an AI model that benefits all
There needs to be a balance between enhancing business’ societal license to operate and the financial Return on Investment from AI. Find out how it can be done.
How Irish organisations can realise the benefits of AI at scale
Machine Learning Operations (MLOps) can help organisations reach their true AI potential. Find out how.