---

Understanding AI Part I: Questions to Avoid AI Disaster

---

---

Technologies like ChatGPT are having a moment. Let’s consider the potential problems and what to watch out for.

Take a close look at the image above, generated entirely by AI (Artificial Intelligence). The prompt was “work-from-home contact center agent” and the image does a great job reflecting that. A typical woman in a typical home office. You can easily see she’s a customer service agent. But gaze at the image long enough and it starts to go wrong. What’s going on with the hand? Faces, offices, computers, even hair… AI does remarkably well. But for some reason, it just can’t seem to get hands right.

And that’s the problem, isn’t it? Results that impress, but still aren’t good enough. If you’re only looking to impress your customers with novel experiences, any AI will do the job. But let the wrong solution loose on your business processes… and it will become a problem!

AI can do incredible things. We frequently recommend AI-based solutions to our customers—and they’re thrilled with them. But it’s not as simple as just adopting any new technology that has this type of capability. AI implementation requires careful consideration and thoughtful application.

What do you need to know? In this article, we’re going to call out the questions—what to think about, what to look for, and what to know. AI can be a messy and complicated subject. While we’re not going to be able to distill it into a simple, easy-to-digest guide—that’s impossible—we will help you navigate the fundamental aspects to consider and the important principles to adopt as you make use of this world-changing technology in your business.

What the Heck Is ChatGPT, Anyways?

We’ve all been hearing a lot about ChatGPT recently. It’s the latest in a long list of confusing acronyms: AI, LLM, LLaMA, PaLM 2… What do these all mean?

Let’s start with AI or Artificial Intelligence. This is an aspirational term, but it can be confusing. We haven’t created C-3PO. “AI” can’t make intelligent decisions. It can’t think. It can only do what it is programmed. This programming is called an algorithm.

One of the most well-known of these new, “AI” programs is ChatGPT. ChatGPT is a chatbot that’s freely available online. It uses a technology called a Large Language Model (LLM). LLMs are an algorithm applied to huge amounts of data with the goal of generating more content that looks like that dataset. This allows the software to understand and generate text which sounds plausibly human. The algorithm is then put through training by humans to tweak its output to sound more realistic. 

There are many LLMs out there. Here are some of the ones making headlines today:

• OpenAI has developed some of the most talked about language models: GPT-3.5 (featured in ChatGPT) and GPT-4 (which powers Microsoft Bing AI chatbot).
• Google has just announced PaLM 2, which will become the engine behind their chatbot called Bard.
• Meta, the parent company of Facebook, has developed one as well. It’s called LLaMA and is used primarily in research and non-commercial contexts.

This technology is about to become ubiquitous (if it’s not there already) because it’s actually quite impressive. These models can process multiple languages and spit out knowledge on almost any topic.

There is a large caveat, however. LLMs are not applying reasoning to generate valid or novel ideas; they’re simply producing text that sounds plausible or believable. So, what does this mean for bringing AI into your business and what are the questions that you should be asking?

#1 Am I Getting Misleading Answers?

Accuracy in AI hinges on two things: the quality of source information and its ability to understand what you’re looking for. Therefore, it’s important to ask a couple of questions. First, is there misinformation that could skew your results? Second, does it understand your input—in other words, is there any ambiguity, sarcasm, contradictory statements, or complex terminology?

Just because AI can generate results that are plausible or sound right doesn’t mean that they’re correct. These models don’t create entirely new ideas or apply reasoning to your request, they just create content based on patterns. That means they don’t take accuracy into account. Earlier this year, Bard, Google’s AI chatbot, made a factual error in its first demo. It’s not just alarming that a chatbot can publish factual errors with such blatant confidence. It’s the fact that it feels trustworthy that’s the bigger problem.

If you are depending on AI to provide accurate answers, the only way to ensure that is to ensure the source data it was trained on is correct. The problem here is that you aren’t likely to get that dataset. The data and the training process are really expensive, so you are likely just going to use a 3rd-party’s trained algorithm and you don’t know what really went into it. We’ll talk about some solutions to this problem in Part II of our series.

Controlling the questions asked is very difficult as well. To take advantage of the technology, you need to let it directly interact with people. That’s the point. It makes it difficult to reduce ambiguity.

#2 Am I Getting Wacky Results?

While AI often generates results that sound plausible, it can also give you something that’s out of left field. These are the kinds of answers that make you question, “What are they thinking?” When the stakes are low, these kinds of errors aren’t such a big deal—we all know the technology isn’t perfect yet. You might even find the errors to be somewhat cute.

But these errors can threaten the reputation of your company and cause people to question your brand. It may be cool for a few minutes to have a chatbot answering customer questions, but one answer that makes the customer think, “What? Why would they respond that way,” can be devastating. Worse, the point of these technologies is to operate at scale. Consider the result of thousands of those moments and what that will do to your reputation. What’s the business impact?

Recently, a lawyer made headlines when they used ChatGPT to do research and ended up citing a number of nonexistent cases. You can imagine the judge was less than thrilled about that and the lawyer may now be sanctioned by the court. But consider what the lawyer’s clients think of him. Worse, think about how all of the publicity around this mistake impacts FUTURE clients. Don’t let your business be the next one making headlines for all the wrong reasons. That’s why it is important to have a professional help you figure out how to apply AI to your business.

#3 Am I Getting Uninspired Content?

Let’s make the expectations clear: when using AI, the results, by definition, will be average. That’s the point of the technology! AI asks, “What is the most common, plausible answer?” And that’s what you get. The results will always be lacking in creativity, problem-solving, and critical thought. Dan Turchin, CEO of PeopleReign, said it best, describing the problematic nature of AI. “It’s the creator economy equivalent of communism—everyone does fine, and nobody is exceptional.”

For example, we prompted ChatGPT to finish the sentence, “The future of AI in business communications…” and received this response:

It might be true and important, but it’s boring… and certainly not useful! An intelligent human can approach this sentence in so many interesting, complicated, insightful, or valuable ways. AI delivers none of these. You need to consider this carefully when thinking about how to apply AI to your business.

#4 Am I Opening Myself Up to Litigation?

There’s a difference between information and information that you can trust. Even if it’s accurate or sounds reasonable, ask yourself: what is the origin of this information? Do I know and trust the source?

When AI delivers unreliable, incorrect, or even dangerous information to your customers, you’re liable. Imagine what it would mean for your warranty if you trusted an LLM and it gave a customer a wacky result. “Plug in the phone, then dump it in a bucket of water.” Now, owing them a replacement product is the least of your problems.

Or consider the question of plagiarism. If you’re not carefully monitoring the source of your AI, you might as well assume that the results are plagiarized at some level. It’s not just an academic question. Do you want to open yourself up to litigation? Are you fine with your proprietary product information or expert content showing up on your competitor’s website? When it comes to your business, it’s not enough to just hope that you’re safe. You need expert help.

#5 Am I Considering the Real Costs?

Part of the allure of AI, and LLMs in particular, is that they’re freely available. Anyone can use ChatGPT or Bing to get AI assistance at no cost. And it’s probably a good practice to start becoming familiar with the technology. It’s free, so use it to help you write an e-mail or put a first draft on paper.

But when you’re thinking about incorporating it as a key component to your business, remember, this technology isn’t without a cost. There has been an enormous investment in developing these models, and those investors are going to want a return. So expect the next wave of this technology to come with lots of advertising and steep subscription costs.

Once you’ve incorporated AI into your business, the costs don’t end there. Live people have to review the results. There’s a training process that has to take place for the results to be valid. This isn’t going to be free. If you want to buy into this future, you have to know what the long-term costs are going to be.

Conclusion

Whether you’re ready for it or not, AI is here to stay—and it’s disrupting business as you know it. But there are giant gaps between what we call AI, what it actually is, and what it might someday become. There’s a lot of work to be done before it reaches its potential.

In our next article, we’ll turn the corner and look at the business benefits of AI and how you can make wise and thoughtful use of it to enhance your customer experience. While there is good reason to be cautious with this technology, there is no doubt that, done right, it can be incredibly powerful for your organization.

Do you need a partner who can help you navigate AI?

The answer is Vertical. We’re unique in the industry because we don’t just sell products; we’re your strategic partner. We’ll provide expert consultation to help you navigate the challenges. Our skilled team will do a smooth and quick implementation with leading technology—so that you can keep pace with competitors. And our support team is second to none when it comes to customer service, knowledgeability, and response time. Our team is constantly exploring and training in the newest AI applications and we’ll make sure you have the tools and information to keep up. We’re always by your side. That’s the Vertical difference.