Sometimes Artificial Intelligence Looks Really Artificial

With the rise of AI writing tools and the ability to create written content in minutes, it’s important to be aware that sometimes Artificial Intelligence looks really artificial – but what can we do about this?

Last week I had to deal with a sensitive AI situation – a client had used ChatGPT to produce a research report. They wanted to save time, so they collated the statistics with some hastily written descriptions and then copied this into ChatGPT with the instruction to “write a management report”.

They wanted to save time, but….the report that ChatGPT produced used terminology and a language style that was so “salesy” and not authentic to the culture of the organisation. I read it and immediately called them as it some of the phrases used so many words it was difficult to read.

I know from my own experiences in the last few months that ChatGPT creates high volumes of words, using 10 when 2 would have done! The result of this situation was that it took longer to edit what the “intelligence” had created than if they had written it themselves.

This illustrates that we need to use these tools with care. Artificial intelligence can be given some wording and will use similar examples and information to create a report but there are many things it doesn’t know. In this case:

• It doesn’t know the likes and dislikes of the audience. For example, it was not able to respond to their dislike of sales style wording – the audience finds that approach pushy and it makes them angry
• It doesn’t know the “normal” style, the types of words and phrases that are in common use across the organisation
• It doesn’t value brevity – it adds rather than takes away. I wonder if it because these platforms can “read” vast volumes of data at speed, and forgets that humans take longer to read!

I am finding that asking ChatGPT to summarise is useful, but if I want it to create I have to tell it specifically to use short sentences and simple words and to keep the volume low. I am very specific because it does not have the intelligence to read my mind – or not yet anyway!