In a recent article, The Age of AI Has Begun, Bill Gates describes generative artificial intelligence as the most significant thing to happen to computers since 1980.
Meanwhile, ChatGPT is stirring up a global storm.
But what does it mean for L&D? Should you use it now? Should your learning technology provider be incorporating it into their platform? Does this mean that robots are coming for our jobs?
In this podcast episode, I speak with Donald Clark, CEO of AI-driven learning company Wildfire, about the future of corporate learning and the transformative potential of generative AI.
Read on to hear about what Donald calls the AI big bang, three ways to integrate generative artificial intelligence like ChatGPT into your learning and development strategy, and thinking about AI as an L&D leader.
But first, Donald brings us up to speed with the advancements in generative AI since the end of last year.
As Donald explains, there are three levels to the history of AI, starting with Euclid in around 300 BC and going right up to November 30, 2022.
The framework for AI has been around since Euclid, who put the first algorithm in his book.
“We had two and a half thousand years of mathematics to get to this point, and there is a huge amount of theory and statistics on probability theory and AI itself,” Donald says. “So, there's that big, long tail and stretch, which is the accumulation of knowledge in mathematics because AI is, after all, math and software.”
However, AI has been hiding in plain sight behind internet tools for the last 20 years.
“When the internet came along, there were several dimensions,” says Donald. “We have to remember that Google search is the big thing here, and the relationship with knowledge changed.”
With search engine algorithms, we can find and retrieve information quickly, and as Donald explains, this is fundamental to learning; who doesn't use Google almost daily?
“We almost forget that AI drives all these interfaces we've been using all the time. There's the 20-year stretch where we've had AI as a hidden, invisible hand behind almost everything we do online.”
We almost forget that AI drives all these interfaces we've been using all the time. There's the 20-year stretch where we've had AI as a hidden, invisible hand behind almost everything we do online.
But then something interesting happened on November 30, 2022: the release of ChatGPT-3.
“The usual quote is that it reached a million people within two days, a hundred million within two months. In five months, this thing is the fastest piece of adopted technology ever in the history of technology,” says Donald. “And I've seen that AI is a big bang, like writing, for ages.”
ChatGPT-3.5 is based on an underlying model with 175 billion parameters, as Donald explains, which is enormous.
“This is the total sum of human knowledge scraped off the internet and used in a socially constructivist manner,” he says. “And then, within a couple of months, you get GPT-4.”
What is essential, especially for L&D leaders, is that the leap between versions three and four is vast in terms of accuracy, guardrails, ethical care, and so on. Crucially, ChatGPT-4 also has an API so that you can link it to other systems.
“And, of course, that's what's happened in Teams, Google, and other contexts,” says Donald. “And now, we have some amazing things happening with Khan Academy and Duolingo where it's massively accelerating the learning process by providing brilliant, what I call, ‘pedAIgogy’ into those products.”
“I think where Gates is right in his article,” explains Donald, “is when you look at the net benefit of generative AI, he thinks that in educational learning in general, it’s massively beneficial. I've been saying this for years.”
But what really blew people’s socks off with ChatGTP, to use Donald’s words, was the simplicity of the interface. “I have felt this sense of release from being lectured to that you get in courses, even in eLearning. And the more I've come to use it, the more I've seen it as this universal assistant,” he says.
The second impactful element of ChatGPT is the sheer breadth and scope of the AI. Although its mediation is a simple, easy-to-use chat interface, it is extraordinary because it replies with a uniquely created answer and is a more normalized form of learning because it involves dialogue.
For example, I asked: In one sentence, what makes ChatGPT-4 so easy to use? “ChatGPT-4's ease of use stems from its highly advanced language understanding and generation capabilities, which enable seamless and natural conversations with users.”
So, where does Donald see generative AI impacting your L&D strategy?
As Donald has found, you can leverage generative AI like ChatGPT by using existing tools that are AI-based, as a universal trainer, for learner support, or for creating content.
First, generative AI will be in all the tools you are using or will use, such as in Microsoft Teams or Google—it will be everywhere.
“Generative AI draws upon a body of infallible knowledge,” says Donald. “It's got some fallibility problems with provenance when it delivers information to you, but this captures everything that's out there and is like a deep mirror, which we don't look at but dive into.”
A second way to leverage generative AI is to use it to impact learner support.
“When it comes to learning,” Donald says, “there is the long-term idea that on the horizon, we have a universal trainer or universal teacher. I mean something that will always be there, and I just ask it something, and it gives me pretty much what I need.”
Generative AI has the potential to magnify the learning in the flow of work process by empowering learners to learn anything at any time.
There is the long-term idea that on the horizon that we have a universal trainer or universal teacher. I mean something that will always be there, and I just ask it something, and it gives me pretty much what I need.
“There is also a big engagement aspect,” he adds. “Hundreds of millions of people are using ChatGPT, but I don't think you've got hundreds of millions of people clamoring for your latest eLearning course on compliance.”
Finally, Donald explains that you can leverage generative AI in content creation.
The significant impact could involve reduced courses, increased performance support, and a more advanced transition away from lectures. There could also be a shift from offering everything as a course to focusing on a demand-driven model as individuals utilize the tools for their own needs.
“There is still a case to be made for sitting back, getting out of that context of workflow for a minute, and collectively saying we are all in this organization, and we need to know this specific information. For example, a product launch or a change in legislation,” Donald explains.
“It might also outlast some of the terrible L&D tropes,” says Donald. “We got stuck in this one furrow with this big tractor plowing away against the grain. It’s not a demand-driven model, which disappoints me because we have a massive skill shortage out there.”
“Things are falling apart because we are no longer training people in competencies, and I'm hoping this will free us from the entanglement with abstractions instead of skills.”
We got stuck in this one furrow with this big tractor plowing away against the grain. It’s not a demand-driven model, which disappoints me because we have a massive skill shortage out there.
So, how does Donald envision ChatGPT or other generative AI helping you as an L&D leader?
“If I were running a big L&D department at the moment,” Donald explains, “I would certainly have everybody under my wing looking at the technology.”
Let's break that down. What do you need to know about generative AI?
“First, you need to understand that it’s creating a sense of urgency, and that's a given,” he says. “ChatGPT is not a fad if hundreds of millions of people are using it, and there is no reversing this now.”
First, you need to understand that it’s creating a sense of urgency, and that's a given. ChatGPT is not a fad if hundreds of millions of people are using it, and there is no reversing this now.
“And secondly, you need to get to grips with it. Don't just ask it a question and wait. Try and refine something, or give it a task or a problem and see how it helps you.”
As Donald explains, every L&D department should think about how they can use this technology now. There's a strategic piece around the use of generative AI in the L&D department. Think about how you are going to position and communicate it at a senior level.
“I would also be strategically looking at your budget overall and saying what portion of the budget we should shift towards doing things more dynamically and providing performance support as opposed to course delivery,” he says.
“L&D needs to move at speed here,” Donald explains, “and hopefully, this podcast might have opened up some doors for people.”
Specifically, he recommends that you take deep dives into any perspectives on AI online now or even log into ChatGPT and ask it questions about how it works, potential provenance issues, and so on. Quick caveat: if you’re using GPT-3.5, it will only give you data up to 2021, so you might not get any contemporary information.
“You can also research ChatGPT through reading blogs or on social media,” says Donald. “Get to grips with it, but do it with a strategic purpose: how will I guide my ship as L&D into these uncharted waters?”
Thanks to Donald for sharing his insights and experience with us! Keen to learn more from L&D experts? Check out our other episode with Donald about five applications of AI models in learning and development or with Peter Sheppard about how skills can only be developed if they are understood in the context in which they are employed.
Want more peer insights on transforming workplace learning? Sign up to become a member of the L&D Collective, and check out our other #CLOConnect interviews with top L&D leaders on driving growth and scaling culture through Collaborative Learning. Or you can subscribe (below 👇) to our weekly newsletter to receive our latest posts directly in your inbox.