Harnessing AI Text Technology – ChatGPT in Test

It’s inevitable that when an unbelievable ‘disruptive’ technology breaks out of the tech developer scene and into mainstream press reporting, the hype is enormous. However, even those who follow the tech scene closely were perhaps surprised at the excitement generated by the latest release of the AI text generator ChatGPT.  

We set out to answer the question—what does this innovative technology mean for your business, what are the pitfalls, and how can you harness it safely?  

What is ChatGPT? 

Simply put, ChatGPT is a language model trained to generate text that feels like it was written by humans. It uses a neural network, a machine-learning process with interconnected neurons that mimics the human brain. By “feeding” the neural network with information—in this case, writing—it learns how humans write and can replicate this skill. Other AI tools aim to do similar for creating graphics, translations, transcriptions, and much more. 

Rather than a person being paid to sit for hours writing a blog post, compiling a report, or coding a website, ChatGPT could do it in seconds, so the theory. 

The initial news stories promised that ChatGPT would revolutionize the way we work. Suddenly, companies were considering if they could shrink their marketing budget and have the software write the copy for their website, journalists were worrying if they’d be replaced, and even coders were being asked by friends if the bots would be doing their jobs soon.  

Use cases for ChatGPT

There is a huge range of use cases for ChatGPT, and which is helpful for your company will depend on your organization, your processes, and your staff.  

A B2C company uses ChatGPT to build customer service bots that make customers feel they are interacting with a human being, improving customer satisfaction. 

A company integrates ChatGPT into its knowledge management system so that staff can access company information easily and quickly, reducing the workload on HR staff.  

A business consultant specializing in change management uses the service to provide virtual assistance to their client’s employees, answer frequently asked questions, and provide links to relevant resources. 

The potential is significant, as organizations can use this new technology to streamline and offload routine tasks. This would free up their staff to concentrate on high-level, creative, or strategic work, making them more efficient.  

What are the drawbacks of ChatGPT? 

In the first weeks after the launch of ChatGPT, and as people started using it, the limitations of the service became clearer.  

It’s a language model, not an analytical or fact-finding model.  

On a basic level, Large Language Models (LLMs) work by predicting the next word in a series of words and are trained using massive amounts of data. ChatGPT was also trained using human feedback so that the AI learned the response humans expected when they asked a question.  

It has a vast knowledge and information database but isn’t trained to evaluate any of it. It doesn’t think or reason, and it doesn’t fact-check itself. When the technology news website CNET decided to have the software write news articles, they found out just how wrong ChatGPT could be.  

Not only does ChatGPT not fact-check its output, it actively creates false quotes, attributes quotes to people who don’t exist, and provides fake research papers as sources. . This doesn’t mean that the software is useless, or that it should be avoided, but it should be used with caution and the output checked for accuracy. 

Ethics, transparency, and governance concerns  

In an earlier iteration of text-based AI, Microsoft hurriedly pulled a chatbot after it became “infected” with right-wing propaganda, highlighting the need for this software to have guardrails in place.  

ChatGPT has filters that prevent it from writing about specific topics, but the data may contain biases and stereotypes that can perpetuate harmful beliefs. 

And that leads us to a question underlying data being used, how it is selected, and who decides what is “acceptable.”  On a societal level, can we trust individual companies to make this decision, and if not, who do we trust? If this is to be regulated, how do we ensure that governments are not creating AI strategies based on ideology?   

On an organizational level, what safeguards can you put in place, and do you have an “emergency brake” if something goes wrong? 

The potential of ChatGPT 

With these caveats in place, there is a huge potential to be exploited in this incredible tool. We may still be a while away from asking it to write complete proposals for a new project, or independently churn out blog posts, but we can take advantage of its strengths while minimizing risks. 

Knowledge aggregation and distribution 

One of the amazing things about ChatGPT is the immense amount of knowledge that it has at its disposal. It’s simply impossible for any human to gather or retain so much information, even on a specific or very niche topic. For example, in January 2023, there were over 4 billion words on Wikipedia. If you were to read every single article, at an average reading speed of 200 words per minute, and read 8 hours a day, it would take you 171 years to finish. ChatGPT was trained on databases from the internet that included 300 billion words.  

And, unlike Google search, ChatGPT presents the answer to a question in an easily digestible format. We see massive potential for companies to enable their staff to access information quickly and easily, for example, via an AI chatbox based on the company training and development database. 

Research and brainstorming 

Use the power of the database to explore a topic. With detailed prompts, you can ask ChatGPT to research anything – from ideas to reduce costs across a supply chain, to how to implement DEI training in your organization. It may throw up ideas you haven’t considered and give you a starting point for the next steps.  

Free up resources  

When you think about using ChatGPT, consider repetitive and time-consuming tasks in your organization. By freeing your staff from these tedious jobs, you can raise their job satisfaction and enable them to tackle more strategic or creative tasks.  

What is next for ChatGPT? 

In February 2023, Microsoft relaunched Bing, their “eternally mocked” search engine, incorporating the power of ChatGPT. Rumoured to be using the newest version of ChatGPT, Bing will combine the capabilities of a search engine with a ChatGPT chatbot. In an interview on the Decoder podcast, Microsoft CEO Satya Nadella talked about the potential of an AI-powered search engine. 

“So, I think we’ve used and evolved to use new tools… I think, yes, some of the drudgery of knowledge work may go away, but that doesn’t mean I won’t enjoy it.” 

In the coming years, these tools will be developed further. Forward-looking companies will promote their use, and look for ways to integrate them into their processes. We would advise organizations to assess these tools, create a framework for safe usage, and regularly review them to ensure they are keeping up with technological advances.

2 Tips for using ChatGPT  

  • Invest in designing guardrails and governance for AI applications 
  • Empower and train your employees for AI usage:  
    • Prompts are important. The quality of the input determines the quality of the output. The more precise your prompt, the better the results. 
    • Critical thinking becomes ever more important. Don’t accept the results as the “truth.” Instead, use them as a starting point to do more research. 
    • Reduce anxiety around job security to leverage the full potential of the human – AI collaboration   
Discuss with us!
Leave us a comment.
You might also be interested in
12
Jan
HR & Sustainability – Shaping the ‘S’ of ESG
20
Apr
Listen
1
Sep
Introducing The Metamorphosis Newsletter