It’s difficult not to seem like an old guy shouting at the moon. This happens when discussing Artificial Intelligence. Let’s not even talk about what wise really means, or doesn’t.

I’ve used a mixture of services, modes and models like Chat GPT, Bing Chat, Dall-E, Gemini and Co-Pilot. No doubt I have missed several. One thing they all have in common is: they’re not quite there. Impressive, certainly. Accurate, less so.

What they are good at is assisting your own research.

None of the services can be trusted to be accurate … yet. Even if they were completely accurate, I would still recommend that the reader performs additional research. This effort can spark new ideas and new thoughts. Use that unique attribute that AI doesn’t have: humanity. Use your humanity to create contrasts, identify paradigms and draw parallels.

As an assistive service, they can be useful to help with tasks from the rote to the complex. GitHub Copilot is pretty good at summarising parts of code in a Pull Request. However, the person who clicks the green button must confirm. CoPilot won’t accept responsibility for approved bad code. If I can’t understand the goal of a Pull Request without using CoPilot, then something is wrong with either me or the author.

Readers of this blog will have noticed additional imagery in the posts. Most of which have been generated using the Microsoft Designer tool, which uses Dall-E. The images you see aren’t great. They often don’t come close to the content they’re supposed to support. However, these are the better images! I’ve passed over several images due to bad spellings. Some images included my prompt as part of the content, such as the words “make the image fun.” Others were so far off the mark that they were irrelevant. I’ve been able to create some free images with quite complex designs. I haven’t thought about hiring a professional artist, so I can’t complain too much. (Sorry digital artists)

It does shine for sparking ideas. I use it for coming up with ideas for holding activities for Scouts, for example. In our current project, we need to present our research around aviation. The content isn’t hard, but the method of presentation could be. Not everyone would be comfortable with standing in front of their Troop. It suggests ideas such as infographics, slide decks, videos and lots more. From this it is easy to jump into other directions, using my own thoughts, experience and humanity.

The impact on our jobs is going to be interesting and I’m not so confident/arrogant to be able to make a prediction. What I can see is that they require a lot of data, they’re expensive to run and they contribute negatively to an already stressed environment. I did have thoughts on integrating “AI” into Taggloo for more natural translations. However, the necessary corpus isn’t available. As a result, any translation could be unreliable and would provide the language a disservice. Performance and cost would also suffer, making it too expensive to be able to run even the simplest of models and therefore unviable.