Customer Experience Customer Service Automation

ChatGPT: The good, the bad and the ugly

By Luigi Loconte 27 April 2023

ChatGPT has been making headlines since it first appeared at the end of 2022. An AI language tool powered by Natural Language Process (NLP), modelling and machine learning, it’s been heralded as a ‘game changer’ for everything from software engineering to customer service.

But, like any new technology, it’s not perfect – and many have been quick to criticise ChatGPT for its problems and limitations.

In this blog, we’ll share the good, the bad and the ugly of ChatGPT to help you make an informed decision about using generative AI in your processes.

The good

There’s a reason ChatGPT has taken the world by storm: unlike previous generations of AI chatbots, it feels like talking to a human.

ChatGPT isn’t limited to responding to pre-programmed prompts. It can understand language in a way that’s almost indistinguishable from human interpretation – and then respond with meaningful and contextually relevant answers.

What’s more, just like a human, ChatGPT learns from previous interactions and has a memory. That means it can carry on conversations without forgetting what you said 5 minutes ago, and it can learn from past mistakes, thereby delivering more useful support.

Where ChatGPT has the advantage on humans, though, is its speed. It can handle vast amounts of data and process it at lightning speed, producing in seconds what would take a person hours. Want to summarise a long document in a few bullets? ChatGPT can do it for you. Want to expand a few bullets into 500 words? It can take a stab at that, too. Want inspiration for topics, turns of phrase, keywords, social posts or emails? In a few seconds, you can have ideas to get your creative juices flowing.

But the real benefit of ChatGPT over other categories of AI tools? Its versatility. It can be trained on a whole host of tasks, from language translation and text summarisation to code generation. There’s no such thing as a one-size-fits-all tool, but ChatGPT definitely comes close.

The bad

While ChatGPT is undeniably human-like, it’s not actually human – and so it does have key drawbacks.

A major one is its lack of common sense. Although ChatGPT can understand and generate text, it can’t reason in the same way humans can. This can result in responses that are technically correct, but nonsensical in practice. Similarly, ChatGPT can’t understand context, emotional cues or nuance. It doesn’t know how its responses are going to be used, so may produce responses that are irrelevant or inappropriate to the situation.

But ChatGPT’s biggest limitation is the risk of inaccuracy. It’s only as good as the data it draws from. If the training data is biassed, incomplete or incorrect, the model will learn those biases and errors and replicate them. This means it can sound authoritative when saying something that’s actually wrong. It can misinterpret inputs or draw on incorrect inputs, meaning you can never rely on the accuracy of what it produces. ChatGPT even has a disclaimer saying that it ‘may produce inaccurate information about people, places or facts’.

The ugly

Many ChatGPT limitations can be worked around if you use it for inspiration and support rather than a complete, unchecked replacement of human work. What’s more difficult to navigate are the ethical concerns.

Data privacy is a key one. ChatGPT uses vast amounts of input for training. You don’t have control over what happens to your data and information after inputting it into the tool. As a result, you should think carefully about what you ask ChatGPT to process – and avoid anything confidential or involving personal data. If any security vulnerabilities in ChatGPT are exploited, attackers could access user data or inject malicious code into the system.

Differentiation is another important concern. By its nature, ChatGPT can only provide responses based on existing information – it can’t create anything new.

Therefore, relying on it for customer experience can lead to generic, un-personalised and potentially incorrect messaging. You can end up reducing the effectiveness of your campaigns because you’re effectively basing your messaging on what others are saying instead of creatively communicating your selling points to your specific target audience.

Another concern is the role ChatGPT could play in so-called ‘fake news’. Because its text is almost indistinguishable from human-generated content, it can be used to spread incorrect information or impersonate individuals, leading to significant social and reputational consequences.

The bottom line

When implemented correctly, ChatGPT and similar generative AI tools have the power to increase operational efficiency and free up human resources for more creative, complex work. But, like any technology, it needs to be implemented with customer care in mind.

If you carefully examine the pros and cons of generative AI for specific use cases and keep security, privacy and customer centricity in mind at all times, you’ll be set up for success.

To learn more about how Engage Hub can support your digital transformation journey, get in touch.