May 25, 2023

How ChatGPT, Bard and other LLMs are signaling an evolution for AI digital humans

From ChatGPT to Google's Bard, LLMs are vastly accelerating digital human technology. Here are five ways.

How ChatGPT, Bard and other LLMs are signaling an evolution for AI digital humans

Everyone seems to be talking about ChatGPT at the moment. Actually, scratch that, everyone seems to be talking to ChatGPT at the moment. 

The New York Times described it as the "best AI chatbot ever released" and within just over a month of launching in November last year, ChatGPT had over 100 million users. By February 2023, its total website visits had skyrocketed to a billion. 

It's not difficult to see why it's experienced such a meteoric rise – ChatGPT is an incredibly impressive piece of technology. It's a language model that uses machine learning to generate human-like responses to your text inputs. 

You write a question, it answers – and quickly. Conversations feel natural and free flowing, with almost no difficulties in understanding what you're asking. Google Bard – as of May 2023 – might be even better.

So, has Web 3.0 truly arrived? Large language models (LLM) like ChatGPT and Bard have certainly captured the imaginations of the entire business world, not to mention plenty of students who think it'll make a great, ahem, 'study aid' for writing assignments. 

At UneeQ, we're extremely excited about the possibilities for ChatGPT, and not because we want it to do our homework for us. Readers of our blog will know we’ve been using the technology for some time. Sophie, our digital human guide who welcomed you to our website today, is powered by such LLMs.

And we believe its arrival has signaled a moment of evolution for digital human technology. Here's why:

1. Brands can save time inputting dialog 

Developing a digital human or AI chatbot hasn’t always been easy. Previously, brands have had to consider all the possible questions their users might ask, as well as come up with a selection of suitable responses. Then, thousands of lines of the resulting dialog must be typed into a backend system, a process that often takes months to complete – let alone update.

That's not all. What if your digital human needs to speak several different languages? Or recognize regional dialects and slang? And do all of this while chatting with users in a way that aligns with your brand and values. It's a big task. 

With generative AI like ChatGPT, companies can craft huge scripts of dialog in the blink of an eye (or near enough), massively cutting down the amount of grunt work involved.

This not only allows teams to dedicate more time to other tasks, but also speeds up the development cycle overall.

Sounds awesome, right? There's a small catch. ChatGPT has a tendency to … fib. For example, we recently asked the platform to tell us a funny story about Sam Walton, the founder of retailer Walmart. This was ChatGPT's response: 

TextDescription automatically generated

Okay, not the most hilarious of stories, granted, but not bad for an instant response! And it would certainly liven up an article you were writing about Walton. The only problem: it's not true. 

Our subsequent research to factcheck the story turned up nothing to corroborate it, despite ChatGPT suggesting it was a "well-known anecdote" that had been circulating about Walton for years. 

Newspapers are already reporting that ChatGPT is inventing media articles that don't exist. Articles that are so convincing, that even the journalists who are cited as writing them aren't sure whether they're real or not

However, ChatGPT is factual and correct a lot of the time. When we asked it to tell us a funny story about Microsoft's Bill Gates, it accurately described the time he was performing a Windows 98 demonstration, only for the operating system to suffer the dreaded Blue Screen of Death on-stage

"That must be why we're not shipping Windows 98 yet," Gates quipped at the time. 

So, ChatGPT can be a brilliant tool for writing dialog, one that can save a phenomenal amount of time. But make sure you have the guardrails in place to prevent too many discussions going too… off piste (more about that shortly).

2. Personality can now be brought to the surface

The best brands engage with their customers by showing personality. It's the MO of every brand Ryan Reynolds has successfully launched. And it's not hard to see why; he's funny, likeable and not afraid to poke fun at himself – people like Ryan Reynolds; and by association, they like the brands he likes. 

That's why emotion-led advertising works so well at the top of the funnel. Almost a third of the best-performing advertising campaigns analyzed by the University of Southern California focused on emotional (rather than 'rational') content.   

For many brands though, it's difficult to build those emotional connections and rapport once customers have shifted from traditional advertising mediums, such as TV and radio, into online experiences of the digital world.

Bringing a brand's personality to life on the (web)page is difficult – can you remember the last time that an online form, FAQ or chatbot resonated with you on an emotional level? Has a 'how-to' article ever given you a belly laugh? Probably not. 

Digital humans seek to bridge the gap between offline and online brand interactions, creating personalized, interactive and immersive experiences. That's what we do best; build lifelike virtual brand ambassadors that really embody the look, feel and personality of your brand. 

Our recently launched GPT module for our digital human platform allows for conservations that are open-ended and on-brand. So, we take the best bits of ChatGPT – it's speed, access to knowledge and conversational skills. But we also improve on some of its weaknesses, like its somewhat stuffy delivery and lack of charm (sorry ChatGPT, we love you really). 

This way, brands can develop digital humans that have unique characters, with drive, motivation, likes, dislikes, and a distinctive voice and mannerisms, all with the conversational skills of an LLM.

Not only that...

3. We can now keep AI on-brand

While we're on the topic, let's talk a bit more about large language models and what else they can do. In lay terms, an LLM is a computer program that has been trained to understand and process human language.

At the time of writing, ChatGPT is powered by an LLM called GPT-3.5, which was publicly launched in March 2022. At UneeQ, we're fondly familiar with these technologies because we've been experimenting with them ever since OpenAI – the firm behind ChatGPT – released GPT-3 nearly three years ago. 

During that time, we've been able to integrate OpenAI's technologies into our digital humans, and we've learned a lot about their capabilities. For one, it's absolutely possible to put up guardrails that keep conversational AI on the right track. 

One of the fears many businesses have about AI is that it can be manipulated to say or do inappropriate things. A Genesys survey found that 28% of organizations were worried that they could face liability for unintended AI misuse. 

AI ethics is very close to our hearts, which is why we developed our five laws of ethical digital human design. And in our time with LLMs we've been careful to ensure that conversational AI stays on-brand, on-topic and steers clear of risky conversations. 

One way we do this is by training the AI to switch between different natural language engines to best answer the questions it’s asked. So, for very specific queries, it might swap from GPT's open-ended discussions to a more closed set of predefined answers. This prevents digital humans going into uncharted territory.

Not only is the AI always learning, but so are we. We can teach it how to respond in ways that best represent brands and their values. And with each iteration of GPT, the platform is only getting better and better at understanding what it's being asked. 

Of course, it has its flaws, some of which we've discussed. OpenAI CEO Sam Altman admits that users should always check any information provided by ChatGPT twice for factual errors. But he also believes it could change the way we live. 

And we don't disagree, which brings us neatly onto… 

4. People are now ready – finally

AI sometimes has a bit of a bad rep. We think Hollywood is partly responsible for a lot of the concerns people express about technology and artificial intelligence. What concerns, you may ask? 

Five years ago, a global study found that 70% of people harbored at least one fear about AI. Of these, nearly a quarter were most scared by the possibility of a robot uprising and the enslavement of humanity! You know, the usual dystopian nightmare scenario. For others, their worries are less apocalyptic (although possibly even more tragic) – one in 10 were anxious they'd get on better with an AI than their real friends or family.

However, despite these fears, the launch of ChatGPT has been embraced far and wide immediately. People just get it. They can see how useful a tool like this can be in their day-to-day lives, and the fact it’s free to play around with at your own leisure helps foster familiarity and encourage some of the innovative use cases we've touched on.

There are still some concerns, as with any new breakthrough. But for the vast majority of people, seeing ChatGPT in action has been enough to understand what an innovative piece of technology it is. Whatever your profession, you'll likely be able to think of ways that ChatGPT can help you in your job. Some even consider that learning how to use these tools could be integral to their career altogether.

And when people instinctively 'get' a technology, that's half the battle. It makes it easier for people to start adopting and benefiting from it right off the bat, because they don't need to be convinced of its merits. The merits are already obvious.

For marketers, innovation experts, customer experience professionals and many more, it’s no longer about “is this worth exploring” but “how can I start using it now!” 

5. There's a whole new world of possibilities 

Perhaps what captures people's imaginations the most about ChatGPT, Bard and other LLMS is the sheer number of fascinating use cases.

Whether its content generation, language translation, creating code or composing music, ChatGPT is certainly a creative's new best friend. But that's not all. ChatGPT can be a cooking coach, a resume writer, an interior designer or a travel guide.

Whatever you might need, it can provide information and advice at the click of a button. There's definitely a spirit of adventure at the moment, as people and brands worldwide experiment with the technology in new and imaginative ways. 

It's also an exciting time for brands who want to stand out because, for marketers, ChatGPT opens a door to a new world of possibilities. It can be used for the obvious stuff, of course, such as generating copy for social media posts or creating eye-catching headlines and detailed product descriptions.

That's just the tip of the iceberg, though. ChatGPT can also be used to analyze data, for example. Simply plug in all your customer data, including purchase histories and browsing behaviors, and it can offer recommendations for personalized product offerings, promotions, and campaigns. Want to generate more leads? ChatGPT can help by writing personalized conversations with potential customers. 

In all these use cases, the technology is there to make people's lives easier, but the human touch is still essential. The technology is there to offer prompts, brainstorming inspiration and launching-off points. Ultimately, though, no one knows your brand better than your employees; LLMs are simply a tool people can use to help make better use of their time, knowledge and skills.

Digital humans, for instance, provide personal conversations that scale to any number of users at once – freeing up your staff to focus on the wider goals for your business. Meanwhile, the brands launching digital humans can save a huge amount of time using LLMs for digital human dialog, instead of doing the data entry themselves.


Wouldn’t it be underwhelming if the potential of Web 3.0 and these incredible generative AI innovations were only about saving time?

Some of the biggest moments of wonder from generative AI have come when tools like ChatGPT message you like a real human; or when DALL-E paints a picture like a real artist. 

The next level for marketing and customer experience is when AI begins to emulate some of the best aspects of real brand ambassadors or customer service reps. When they make you feel heard, understood, engaged and well-looked-after.

That’s why we’re so excited about what tools like ChatGPT mean for digital humans like ours.

So, what’s next?

It's still early days for large language models and generative AI. But in the short time it's been available to the general public, there's been a massive amount of interest in what AI can do, and we expect to see even more exciting use cases over the coming months. 

We’d be lying if we said we predicted how popular ChatGPT would become. We've been using the underlying technology for some time with our clients, and have been singing its praises for years. But the sheer speed and scale of its launch into public knowledge has been truly remarkable.

Now feels like the start of Web 3.0.

When the AI doesn’t just move like us or think like us, but when interacting with one face-to-face feels as fun, friendly, immersive and emotionally impactful as it does speaking to a real person.