I love experimenting with new tools. When generative AI broke into the mainstream with the release of ChatGPT in November 2022, I was ready to get stuck in! Luckily, the culture at Accenture, my previous employer, encouraged experimentation and figuring out what works, within a broad set of parameters.

Here are three ways I use OpenAI’s ChatGPT and Dall-e Antrhopic’s Claude recently, while preparing for a keynote speech on the future of international development.

1. Identify whitespace. By being able to give you summaries of the existing body of work on a particular topic, you can also identify what is not being discussed as prominently and therefore might be a lead for a gap in the conversation.

   
   

2. Suggest word combinations. Large Language Models, have access to countless words within the context of a sentence, and are thus able to create possible word combinations, at a faster pace than I could. When I was writing a sentence for the speech, but it was not quite hitting the mark with what I wanted to express, I would use the LLMs to get to alternative phrasing.

   
   

3. Create images. Dall-e has been a revelation. I used it to test ideas or create my own images that illustrate the content of my presentation in powerful ways. On a separate note, I also use it frequently for my LinkedIn posts. When writing about my son’s comment that "girls like unicorns and boys like football", I asked Dall-e for images to accompany my post. The prompts were: ‘a boy playing happily with a toy’ and ‘a girl playing happily with a toy’. It returned stereotypical images, reflecting inherent biases in the training images.

What these tools have not been able to help me with (yet!)?

In a nutshell: Coming up with distinct and differentiated ideas. When I first started the speech prep I asked for inspiration on the future of the international development sector. It gave me back content along the lines of ‘sustainability, community, technology’, which are reflective of the conversation today but not how this might evolve or change in the future.

I experimented with various prompts, giving the tools further context and ‘characters’ but ultimately I realised they are useful for topic research but not for creating new insights or ideas. Instead, I listened to webinars and brainstormed with friends and colleagues, who within minutes shared sparks of inspiration.

Undoubtedly, generative AI has huge potential, but so far, it’s another sophisticated tool in the arsenal of futurists, researchers and professionals, than a replacement for human ingenuity.

Yet, many questions remain for organizations.

As businesses increasingly adopt Generative AI and build their own enterprise-wide large language models, how can they ensure that they don’t reinforce existing stereotypes that are already visible in such models? And how can they do so, while allowing their people to benefit from the potential of Gen AI?