Should I be using AI in my business?

This is becoming a common question that we're asked as part of our consultancy engagements - so, as the forward thinking business that we are, we thought we'd get ChatGPT's opinion on it:

I mostly agree with what I got back. When discussing with clients, i'll generally explore areas that it may be of benefit, but also highlight some of the drawbacks. That usually results in them asking me why I don't like AI and if I'd ever use it, and the answer is that I do. I'm a big fan of Midjourney for creating banner images for our local Defcon Group site, https://dc151.org when stock images don't quite fit with the theme, and sometimes I'll use ChatGPT to assist me in writing the first draft of an abstract on a consultancy report to shake off the writers block (while being careful in my prompt generation not to disclose any customer information).

I believe the biggest risk of AI use in it's current guise is it's accuracy - too much reliance placed on something that has a history of being "confidently wrong" might be a mistake. We, as humans, should be diligently checking the output to ensure it is correct - after all, there are definitely 3 R's is strawberry, and Google's efforts to use AI to answer simple search terms sometimes produces some misleading results:

(for the record, we're not a Dynamics 365 consultancy company and there's no mention of Dynamics on our website so this is a pure hallucination).

Another issue might be localised training data, and the mangling of different data sources by LLM's. This can be countered by very specific prompt generation, but still a risk if you're blindly trusting (and making commercial or legal decisions based on the output). It's important to know that LLM's are effectively a regurgitation machine - they will, for the most part provide output that is directly derived from their training data. It's also worth noting that by default, data and prompts that you input into ChatGPT and other AI platforms may subsequently be used as training data, and could potentially inadvertently result in your confidential data or intellectual property data being shared. OpenAI however do have an option to disable this.

If you're going to use generative AI in your business it may be worth investing some time in training yourself or your team on effective prompt engineering and to be very wary of the data that you feed back into the machine.....