How AI Can Help Your Business Thrive: An Interview with an AI Developer Proving AI Isn’t Scary

As Artificial Intelligence and Machine Learning technologies advance, businesses must explore how they can best leverage these technologies to increase growth. To comprehend the power of this technology, we connected with Mark McKelvey, AI/Machine Learning Developer for Stacked Analytics, to uncover the most current trends in AI/Machine Learning and why organizations should not shy away from investment.

What is Chat GPT?

Stacked Analytics:

I think of ChatGPT as a very broad chatbot.  We’ve all interacted with chatbots on a website or call customer service and they try to prompt you through an answer.  ChatGPT works on a much wider subset of information, so you can ask it anything.  It’s been trained on a wide variety of text, versus specific customer service content that might only apply to a specific use case.  So, it’s a smart chatbot that’s been trained on books, webpages, code repositories, etc.
There’s a ton of data that has gone into natural language processing algorithms.  The data science community started training these models and they just keep improving.  Open AI took it a step further and hired a bunch of people to improve the output. So, the model is trained on all this information, then the programmer would start asking questions, and grade the responses or provide a better response.   Now a model exists with much better training data from humans answering questions and feeding that back into the model for improvement.
I heard the dataset contained 175 billion columns. 
Stacked Analytics:
The model is reported to contain more than 175 billion parameters.  Parameters are how neural networks make decisions.  So your question gets turned into a series of numbers and the model has 175 billion decision points about the series of numbers to produce the response output.

So why should local businesses care about AI and Machine Learning?   Why has this suddenly hit the mainstream?

Stacked Analytics:

Because it works really well and it’s free.  Anyone can go to the website and type in a question and see how it works. You don’t need any kind of technical prowess to use the tool. It’s simple – here’s a website with a question prompt.
And it’s better than anything we’ve had.  People use machine learning all the time and they never notice it. It’s the first “machine learning” branded thing with very fast adoption. But you know, when you go to Amazon or a Google search or anywhere else, you’re using these kinds of models.  Every time we use a chat bot, we’re using natural language processing, although it can be, you know, poor in many cases lol…!

What’s next for AI?

Stacked Analytics:

Well, I think the models that ChatGPT uses are very good and we have very good vision models as well that are good at identifying images.  The machine learning world is cool in the fact that the underlying models are available to use.  Some are pay models, but in other cases you can download the model and tweak it to your needs for no cost, along with the code that’s used to produce them.  ChatGPT expensed $10 million in Microsoft Azure credits in order to build that model.
Most of us don’t have access to create something like that, but we can use the already trained model.  At this point in time for businesses, the models are good enough now that you can trust them. You can expect that the chatbot experience will be better than what it has been in the past.  It could help with a user experience that you might be proud to put on your website and won’t frustrate your customers or your potential customers.

Can you share a few simple AI business use cases?

Stacked Analytics:

A few quick examples we’ve been approached to work on include:

• Call Recording:  We take an audio file of an inbound call and use AI to “Tag” it with the result.  Did the call result in an appointment, wrong number, current customer service call, etc.  Many businesses track calls, but few classify the result to understand ROI, because it’s very time consuming.  AI can help dramatically speed up this task.
• E-Commerce Store:  Use AI to quickly narrow down options within an online store based on budget, style preference, availability, etc.  Most stores use “filters” but those can be very cumbersome – this streamlines the experience to quickly serve highly relevant options fast.
• Internal Sales Training – use AI to “listen” to any numbers of calls and it will record sentiment such as positive or improvement feedback for your sales team.
• Customer Review Sentiment for Multiple Location Business – use AI to quickly understand detractor categories.  Companies use surveys and review platforms (TripAdvisor, Yelp, Google Reviews) with free text fields, and classifying the cause for complaint is historically a tedious manual process.  We can use AI to do this categorization for us.  For example a restaurant might find that one of their locations get a lot of complaints about cold food.

Many businesses have no idea how much technology costs to build.  How should they financially plan for investing in AI?

Stacked Analytics:

I would approach financially planning for AI in the following steps:

1.  Think about a service or business task that is repetitive and could be solved with AI – Let’s say it’s taking customer service calls inquiring about delivery status of an order
2. What tools and technologies could be used to solve this problem?  Are there solutions out of the box or does it make sense to build our own?
3. How much does it cost to employ the labor to field those requests?
4. Seek engineering or IT and get a development quote
5. Evaluate cost/benefit

There’s no reason price should be astronomical to implement some of this technology because most of the work has been done. You will have some ongoing cost to use their APIs, but it’s very low for what you’re getting, in my opinion. We’ve been asked to do these projects for a number of small and medium sized businesses with pricing agreed upon.  There’s value to explore.