How Charisma.ai is changing the game for the video games industry

Charisma.ai’s CEO Guy Gadney was interviewed by PocketGamer’s Dave Bradley for the first edition of his AI Gamechangers newsletter. Sign up for the newsletter by following the link, and read the full interview below!

Behind the scenes: the Charisma.ai story graph.

AI Gamechangers: As we sit down to chat today, there’s news that video game voice actors and performance artists have called a strike, just as Hollywood actors did, over contract negotiations focused on AI-related protections. This is going to rumble on, isn’t it?

Guy Gadney: The thread that runs through this, and always has done, is ethics. It’s not just a beard-stroking, “nice to have” thing: ethics are quite fundamental to the way these technologies need to be designed.

When we were pitching Charisma, back in 2018, to a high-profile investor, we told him, “We're writing a paper on unconscious bias and AI datasets”. And he said, “That sounds like a distraction.” Needless to say, we didn't do any form of commercial agreement! For us, it is very important that the human is designed in from day one.

I've long quoted the line from Jurassic Park, “Scientists were so preoccupied with whether or not they could, they didn’t stop to think if they should.” I'm not surprised that unions and people around the creative industries are querying how these systems should be created.

We're starting to see the downturn of early AI companies that had huge VC backing back in 2022. My sense is that was unadulterated tech hype, fuelled out of the Bay Area. VCs falling for the hype. A number of those have gone very quiet. There are two reasons why. First, they didn't include the human in the mix! They came out of “big tech”, not out of the creative industries. So they were thinking in a very pure tech software way, rather than actually thinking about the end user, which is just the most insane way of designing a product. And secondly, which is linked to it, they didn't think about what benefit it was offering to games, or to storytelling, or to any process. They just did it, because they could. That breaks a rule of venture capitalism: what's the problem you're solving? And if you're creating a problem to then solve it, that’s a cardinal sin. So I'm not surprised game creatives are putting up resistance.

AI Gamechangers: With your emphasis on the human element in AI development, how do you think we can better bring technologists and creatives together to work effectively and ethically in this space?

It is very important for all forms of the creative industries (including publishing, advertising, film, TV, games, everything), not to sit back and grumble but to actively engage with the various technologies that are out there so that an informed opinion can then be had. There is still an enormous amount of fear about AI. And to my mind, that fear is born out of a lack of understanding, a lack of education, about how these tools work.

We run workshops with writers and other creatives. And at the beginning of the day, somebody will say, “I'm never gonna touch it! I'd rather put a gun to my head than use AI.” But by the end of the day, they'll say, “Okay. It's not as existential as I thought.” We often put up a slide with AI code, as a spoiler alert, saying, “AI is not creative.” Of course it’s not, it's a machine.

As humans, we have an innate fear of the unknown. It goes back to our fear of the dark. However, it's important to know what is actually a risk and what is not. It's fine to be cautious, but if you're overly cautious then you have stasis, and that doesn't help any form of development or evolution at all. What we try and do is switch on the light on the bedroom light and say, “Look, it's okay.” And I don't mean to sound patronising. But it is a critical moment in time where this needs to happen. 

Look at countries like Finland, which had its first national AI strategy back in 2017 – that was about educating the entire country, regardless of whether you were working in a restaurant, road-building, blue collar, white collar… Everyone needs to know about AI and understand what it is, so the country can step forward. At a national and indeed international level, it's very important that we have education so we can make informed decisions about not just whether to use it, but where to use it in our own pipelines and our daily lives.

AI is such an interesting topic because it encourages these conversations. We rarely had these conversations back in the mid-'90s or even in the 2000s. We didn't talk about the ethical implications of building a website (although funnily enough, we did a little bit when I was at Penguin Books!). We never had the level of self-inquiry that we do now.

AI Gamechangers: When you spoke on stage in San Francisco in March, you mentioned that AI's biggest sin is that its output can be boring. Should this be a bigger worry for the creative industries than any fears about it being off-brand or inaccurate?

Guy Gadney: It comes down to where the guardrails are. And for us, the Charisma AI guardrails are our foundations. We're able to start with scripted character dialogue and scripted storytelling and then let out the reins where improvisations in specific moments can be deployed through generative AI.

The problem is that other well-known AI systems operating in this space were built on large language models, which have no guardrails. So they then hurried to try and rein in the horses that had already bolted. That's hard. What you're seeing in the evolution of ChatGPT is that it has become more boring! The fear when generative AI first came out was that it would say something racist, sexist, homophobic, misogynistic, all of these terrible things and that Disney’s share price would tank if they used it, and so forth. But in reality, it blocked any form of interesting characters. Imagine the entire Star Wars canon where there is no Darth Vader. That would be terrible. No antagonist, no sense of engagement or good storytelling. We've run into it sometimes. We were working with Warner Brothers on the character of Steppenwolf from the Justice League film. Steppenwolf, as fans of that superhero franchise will know, works for Darkseid, who is trying to eradicate and dominate planets. We put that biography into Open AI to test… and it breaks it, flagging it as genocidal. Okay, so how nice would we have to make a character like Steppenwolf or Darth Vader?!

That's where we are now. I see an evolution to small language models rather than large language models, which can ring-fence evil characters and allow them to come to life in ways that are good for creativity.

We started from guardrails, and that's really our secret sauce: our ability to give writers and narrative designers that power. We then allow them to deploy generative AI as a paintbrush, a particular technique, rather than building the whole thing on top of it.

A still from our latest AI-powered video game, Project Electric Sheep.

AI Gamechangers: As AI becomes more prevalent in game creation, do you think player expectations are going to change? Do you think gamers will want more flexibility and interactivity in their gaming experiences?

Guy Gadney: I do. But it’s this two-sided coin of AI: one side is productivity and efficiency, which is where all the studios are looking. “How can we do this faster? At lower cost? How can we shorten production timelines?” The other side of the coin, though, is innovation. “What can we do that we could never do before?” That part is interesting because it opens up opportunities for gamers who are frustrated with today’s companion characters, NPCs who just say, “Here’s a new mission,” or “There’s something to see over there.” It's like Clippy for games! So frustrating. But imagine writing characters so you walk into a bar, and it’s populated Westworld style, with 30 people, and what you say to those characters changes the entire dynamic of that scene. That's cool because you are now inside the experience – you're influencing it. The characters are well designed and what you do has proper agency.

The problem we've seen today is a fundamental games industry problem, which is that as a gamer, you don't have proper agency over the NPCs. They're not pushing back enough. I think that's what we'll start to see coming up.

The whole point about Charisma as a platform was to start from a principle of good storytelling and good narrative design, and then build the NPCs. Don't build a chatbot and then try and wrap a story onto it! That’s not going to work. Start with a good character design. Then you've got agency. For us, it's about character, it's about the NPCs. it's about storytelling. If you're interested in storytelling and the characters, then that's where we come in.

AI Gamechangers: To your first point, many game developers are pondering how to use generative AI in their production pipelines. What advice do you have for studios? Is there an obvious way in which modern generative AI can slot into a game development process ethically and productively?

Guy Gadney: Let's roll back because AI is a massive, all-encompassing term! You have many different types of very distinct technologies. “AI” is already being used in game development pipelines, because you've got autocomplete in Gmail. From there it goes all the way up to generative worlds in massively multiplayer metaverses, and everything in between. The way we advise companies in this space is this: look at your pipeline. Look at where the pain points are. Look at what you’re trying to achieve, but can't currently achieve through regular technology or other limitations. In some ways, it's a very simple question to ask, and goes back to, “What's the problem you’re trying to solve?”

AI includes the ability to have natural language processing, machine learning, or the analysis of unstructured data, perhaps data on how players are using the game, generating feedback on improvement capabilities. There are so many different ways. If you’re planning a game that has just been put out there and is not going to be touched after it's launched, then don't bother with analytics because you're not going to touch it. But if it's an evolving, persistent world, it's fundamental to the way that product evolves.

I’ve sometimes said that AI is the cause, and automation is the effect. Study the effect. It's much better than getting caught up by bright, shiny things and feeling overwhelmed by the pace of change. If you look at automation, you start to query your own processes. You can think about how it may be useful for you. And I don't mean automation in terms of replacing people, I mean automation in terms of making your job easier and better. We've done that since the invention of the wheel (and probably prior to that!) so in some ways this is nothing new. The games industry is a technology-heavy industry; it makes sense that automation and technology go hand in hand.

AI Gamechangers: Charisma was recently at the Digital Hollywood event. What was the thrust of that? Can you talk a little bit about the connection between AI, games and the film industry?

Guy Gadney: At the Digital Hollywood event, we discussed two things: business models and how they are evolving in entertainment generally, and digital humans and their evolution and different use cases.

I'm quite broad in my own curiosity about different industries and seeing where the parallels are. When we first set up Charisma, we thought it would be a slam dunk into the games industry – makes complete sense. From a business model perspective, it's about engagement, longevity of gameplay time, reuse of assets, and all of these things. As time has evolved, we've realized we're a much more agnostic creator platform than maybe we'd originally thought. We've powered immersive theatre productions in London, we've powered VR training modules for disadvantaged kids in LA, we've powered language learning, and all sorts of things.

When we ran our early writers' labs with Charisma, we welcomed writers from games, film, TV, theatre, poetry… And the ones who adopted interactive narratives faster were in immersive theatre rather than games. I was surprised by that. But I sense it's because game writers weren’t used to writing dialogues – they were used to writing monologues. I mean that in the sense you walk up to an NPC in a traditional game, bump into them, and they’ll say, “Hey, buddy, watch where you're going!” Whereas if you were to have a line where a character walks up to you and says, “Let me tell you a secret...” and draws you into the story, now that NPC has something in its personality, that I want to be part of. It's a very subtle shift. But it's a shift that immersive theatre writers are more used to because they're used to the fluidity of the story. It’s more fluid than simply a branching narrative.

A screenshot from the Mission Boat scene of our conversational AI game, The Kraken Wakes

AI Gamechangers: A year ago you launched The Kraken Wakes as an AI-powered game. What lessons have you learned from that experience?

Guy Gadney: We learned so much from that. The Kraken Wakes is a classic book by John Wyndham who wrote Day Of The Triffids. I love the book! I think the characters are really great. I wanted to meet them, they're interesting. That was the catalyst for working with that title – I felt Charisma could bring it to life in a way that no other system could. 

It was also a way to see how audiences and gamers responded. It has five hours of gameplay, and you’re properly immersed – voice in, voice out, you’re speaking to the different characters. You're recruited, as yourself, to be a journalist covering this alien invasion. We were surprised, I suppose, by the bad behaviour of some of the players! And how really quite far over the line they went. We can see what people were saying in the conversations they had. Fortunately, we can bring it back, and we can kick people out if we want to. That was interesting.

I think adapting a classic novel like that is a really high mountain to climb. It was a very ambitious project. We ran it for over a year online and have since taken it down because we want to focus on other areas. As with any live and dynamic game project, it required a lot of ongoing support.

It was very much a narrative-driven game. It wasn't that it had many multiple endings; it wasn't what you did in the game that was important, it was how you did it – by the end, it's how you left the characters and the relationships that you had. The planet will always be invaded by aliens, you will always have to escape London and go to the cottage in Cornwall, those are key events that will always happen in that title. But at the end, when you're reflecting on everything that's happened, that is infinite, because that’s what we bring you with Charisma. Ironically, it was a project ahead of its time. We learned a lot, and we'll revisit it. I think we need to probably simplify it a bit.

AI Gamechangers: What’s next for Charisma.ai?

Guy Gadney: Charisma as a platform was developed seven years ago. So we were pre-OpenAI! We've carved a very strong niche in our particular area. Funnily enough, we're seeing a lot of growth in companies that have tried using large language models to run characters in their games and are finding that they are not delivering for them. They are expensive, they take a lot to set up and maintain. So we're winning those contracts now. People are thinking, “Hang on. Charisma has solved this already.”

We're starting to see the maturation of people's approach to AI, and as a mature company, we match that. We’ve been experimenting with generative AI since GPT-2. And our belief is that there is a lot of opportunity there. And that will be in the nature of how we can grow the creative industry of storytelling, and bring new people into it, with new tool sets and new capabilities. We're looking at how we can mushroom what we do, and have a much broader impact on the industry. It is an ambitious goal, but we're working in stealth at the moment on a new research piece. I think it will have a massive impact on the creative industries across all sectors, including games, and on how narratives can be created faster, and how new forms of storytelling can be pioneered, without the risks that currently exist in the media industry, which is where the money is.

From a vision perspective, I think we've seen an enormous move of revenues away from the games industry, from publishing, from the media sector, into big tech. Just look at the move from advertising away from Channel Four or whatever into Meta; they know it's an existential crisis. So if we can create an ecosystem that can power the next stage of the creator economy, then that's a good thing. We're working on that at the moment and hopefully breaking surface later this year.