AI and how we use it
The explosion of AI has been something to admire over the last year or so. It shows incredible potential but, being in its infancy, still has potential pitfalls.
One thing to note, are the amount of companies advertising their AI capabilities which have been customised to their market. Much of what we are exposed to is hype and it’s fairly easy to tell when this is the case. When AI isn’t used but marketed as such, you can tell fairly easily that you are being scripted limited choice and the quality is often bad. Where AI is used, companies should be training AI models in expert knowledge for their field of interest. They then, deploy their own AI model or create an interface to something like Gemini or ChatGPT. A good example of where this happens is the Bing search engine.
Bing, is a good example of where you need to be wary. Bing uses ChatGPT technology and is a big backer of OpenAI (who produce ChatGPT). However, something strange is happening. Bing will often produce results which are very different to ChatGPT. I don’t know why this happens but you will often get results in Bing which are incorrect. Not just wrong, but it’s as if the AI is hallucinating its results. The specific example I have in mind is Companies House data. I don’t want to pick on Bing as for many people, including myself it provides an interesting and different way of getting the information you need.
What information you’re presented with is something you don’t control unless you give specific instructions. This means sources and how it interprets the content can be open to question and always has to be checked. AI now has the habit of presenting links to sources for you but even these links can be incorrect. There are numerous articles which tell of horror stories where companies and individuals have failed to take precautions and jumped head first into a situation which cost them their job and their firms reputation.
Given all the dangers, what can you use AI for? Some ways we use are the following (but we’re expanding its use all the time):
- rewording of content. We always produce our own content but we often use AI to see how it would present information for a particular market. We compare and update our content as needed. The reason this works, is we have created the wording, giving context and limiting the information AI looks at.
- creating graphics. This can be very frustrating as AI can get fixated on certain styles or content and no matter what you ask for, it insists on doing what it thinks you want. But… it sometimes works and we use it to gauge the progression of AI. How it understands content and imagery are easy to measure.
- summarising large amounts of data. Again, this needs to be checked each time but it’s much quicker doing quality control than summarising. AI can be brilliant at consuming vast amounts of data ( such as sales reports) and produce a summary.
- getting suggestions. Sometimes you may need to learn about some topic quickly or you’re looking for inspiration. The training AI does, exposes it to huge amounts of data. Use this to your advantage.
- creating programming code. You can describe what you want, in the environment you work with and the AI will generate all the necessary code.
These are just some basic ideas. Getting AI to complete simple tasks is probably the safest bet (the repetitive work). This allows you to focus on being creative or focusing on more important decisions. AI is developing fast and the latest versions are much better at understanding context. What they don’t do is ask you for more information it thinks it needs to give you a good answer. Hopefully, this will come with time.
Recent Works
