What is AI and is it useful for the small business owner?

Part 1 – A General Overview

I’m sure that you’ve seen all the news articles about the current ‘AI revolution’ and how it’s changing the face of everything from art to medical science. You might have even ads or gotten spam mail shouting about how AI is revolutionising business. Perhaps you have noticed that all the search engines and business software have now got AI modules in them, whether they need them or not. The question is this: is it going to be another big tech bubble that bursts all too soon, like the Dot.Com boom of the 2000’s or the WEB 3.0 of the last couple of years, or will this be a useful tool for the small and medium businesses of New Zealand like you?

So what are these AI tools?

The tech behind it is very complex and far beyond the scope of what we need to cover here. On the surface level reading, these tools are nothing more than the same systems that you encounter everyday when you use your smartphone to type in a text message and what the phone predicts will be the next word in the sentence. What makes the difference is the scale of how it works. None of these systems are Artificial Intelligence, even though everyone uses that term. What they are really is ‘Large Scale Learning Models’; where they use the same prediction methods but instead of making ‘guesses’ of what the word will be based on weighted averages of what words links to the current one based how a language works, they take millions/billions of web pages and build up massive predictive models of whole paragraphs or pages based. Back when I was studying Information Technology, we covered this under the terms of Knowledge Bases, where you built a network of answers based on the questions a user might ask. The more questions and the more variations of those questions the more accurate the answer. The current Large Learning Model instead uses the vast amounts of information on the internet to build these relationships on the fly so there is no hand coding. So the tech isn’t new, just highly refined, using vast data sets, and automated to the point where no direct human interaction is needed.

So where are we going to encounter them?

Currently everywhere. This goes beyond just the direct ‘chat’ models and AI art generation that are grabbing the headlines. Microsoft has already rolled out the technology underlying ChatGPT (currently the most well known and successful example) as part of their Office Suite to allow users to have the application generate suitable equations to process spreadsheets data, and add additional grammar correction tools to word processing. The three big search engine companies: Google,Microsoft and Apple, all added and expanded their result systems to not only generate better results, but to create new content summarising the content of many source web pages.

We are also already seeing various companies using these systems to automate their customer service side of their business. Having AI models directly interacting with customers to deal with the bulk of the general queries online or over the phone, and only passing it on to a live person when the model reaches its limit or hits hard coded points that must be dealt with by human interaction. It’s easy to predict that the larger companies will invest more and more into such technologies to remove the most ‘troublesome’ part of their operations, staff.

So what are the downsides of using AI software?

Like all technologies we are seeing companies racing in to add it to their systems without considering the long term consequences of what that will mean for both their businesses and their customers. Again this is a case where the hype far exceeds what the current results can justify. Especially considering how AI has several issues that will take a lot of work to overcome, or might never be fixed.

The largest problem is AI software can be incredibly wrong in its output. The term that experts uses is ‘hallucinating’, where the model generates results that are faulty. Theses can come about from the systems using bad or malicious sources, errors in the prediction model, or even hard coded into the software due the creators own biases. What makes it worse is that these hallucinations are expressed with such ‘sincerity’ that a person who doesn’t have a basic understanding of the subject could easily believe they are true and don’t follow up with independent research. There are already several published cases of these hallucinations causing problems for people in the real world. One recent example is of journalists being asked about content or cited for articles they never created simply because an AI model reported the journalists had. Another example: a teacher failed a class (depriving them of their graduation) for using ChatGPT to write their assignments by using a different AI system to analyse it. It came out that of the few cases where students did use ChatGPT it was only to generate sources or first drafts and create their own original work from the results. None of them had used any form of AI software to create the assignments handed in.

Another issue is the ethical behaviour of such models. All of these AI models rely on scanning through hundreds of thousands of sources, often without permission or payment for use of other persons or businesses copyrighted material. AI Art has been the biggest storm about this where you can use AI systems like Stable Defusions to create new images in a style of an existing artist and refine it to the point of being almost if not identical to the artist’s work. Beyond the issues of use/‘theft’ of the artist’s work, it could also make it easy for criminal activities such as fraud or using it to misrepresent the original creator.

So what does this mean for small and medium sized businesses?

The reality is, like other revolutions before it, now that the genie is out of the bottle and we have to deal with it. Business large and small will make it part of the tool set they use to run their business work. Even if a business chooses not to use AI models, they will have to have at least a basic understanding of it because these systems will effect their operations in many different ways over the coming years. So where do I feel that SME will find the most use for the tools of AI? I see it serving in the roles of marketing or content generation and to automate smaller tasks that all businesses need to do.

New Zealand is a nation of sole traders and small businesses with less than 25 staff. I can easily see that for many of them having an AI system managing the admin will be a worthwhile spend. Systems that can easily manage online or even over the phone appointment booking. Tools that serve as the customer portal to answer queries and sort through emails. Automating the processing of invoicing customers and generating orders to suppliers. As long as they are carefully set up and reviewed on a frequent basis to make sure they are still fit for purpose then they will be a useful addition.

Another area is the creation of content for marketing purposes. The reality of the internet driven world is that to stay relevant on searches and to keep up with the big companies social media, you need to generate content of some kind. Content that you the business owner might not have the time or skills to generate yourself. (Though we have covered this problem in previous articles and given lists of useful tricks to make it easier.) These AI tools can be the way you can stay at the top of the search engine results by letting you generate the content that the engines like to see while keeping the workload to a minimum. In fact, in part two of this series I will be putting that to the test by getting Microsoft’s Bing ChatGPT search engine add-on to generate an article for us examining the advantages and disadvantages of using AI tools to generate marketing content and adding my own thoughts of what it generates.

So this has been the overview of what the new AI revolution is about, how it works, what it can do, what risks it might have, how a SME could put it to use. Part two of this article will follow the second part of June where we put ChatGPT to a test to see if it really would be useful for your business needs.