What is the future of art without artists?
One of the most common questions I receive from players of our Riskgaming scenarios is whether I can use AI to produce more of them. Ambitions vary. One player asked if I could double the number of games I publish each year. Another thought that, with AI, I “could probably publish 1,000 games per year.” Last year, we published three scenarios. Looking at my editorial calendar, we have three prepared for the first half of 2025, all human-crafted.
Like most creatives, I used to take great umbrage at what has become an inevitable question. You, the player, have just had this extraordinary experience, and the response to that unique moment is a desire to commodify it and rip the game designer (i.e. me) away from their creative endeavor? Slightly suaver players at least preface the question with a compliment before heading for the jugular.
Over time, I have come to accept the question more positively — people loved something, and in our consumerist American mindset, we want as much of that something as possible. There are more than 8 billion people on Earth, there are innumerable Riskgaming scenarios that can be devised, and I and my small team of independent designers can’t possibly cover them all. What’s wrong with a bit of intelligence-driven mechanical reproduction?
For one, I have experimented with AI for game design, and the results just aren’t ready yet. A lack of datasets (Riskgaming, wargaming, board gaming and more) is clearly a major hindrance, but the bigger challenge is really one of taste: what makes Riskgaming special is that I know my audience, which means I can play with them like a puppeteer works a marionette to teach them a series of intentional lessons. I know the blind spots of most of my players, and I can implant them in specific scenes to trigger a specific psychological effect (dopamine or cortisol, depending).
Those scenes are carefully arranged; the narrative behind every Riskgaming scenario is its thesis. Like a good novel, the ending shouldn’t shock by surprise, but rather shock the reader out of stasis, opening them to a new way of seeing. That requires context and empathy with the player, and that’s not something I have been able to get AI models to understand, even with extensively written prompts.
This approach also dovetails with my view that great works must have single authors. Unlike the worlds of science and politics, where almost all products begin with teamwork, art is rarely co-bylined. Even culture produced by massive teams — such as films and video games — have authorial vision as a key constraint. Films mostly have sole directors (as required by the Director's Guild of America) while video games typically have a single lead creative director. I take the same approach with our game designs.
Yet, the medium as I have constructed it has its limitations. Scenes generally have to happen in a pre-set order, and there can’t be narrative branches that change the game’s path. The choices in each scene have to be limited, for their effects have to be computed back into the game’s state, and there are limits to the complexity I can code. Thus, in Hampton at the Cross-Roads, a player has the opportunity to host a press briefing with someone, ranging from Joe Rogan to The Guardian. But there are only seven options available. Or in Experimental Automata, all policy debates are circumscribed to just three possible options — a very tight Overton window.
I am skeptical that AI can fully understand taste and the qualitative aspects of art, if for no other reason than knowing what’s in your audience’s head is untrainable. The few people who publish their thoughts are so skewed (and they are likely even lying to themselves) that I can’t fathom how AI models could gain a theory of mind for the diverse people who inhabit society.
What AI will do quite soon is degrade singular authorial vision into something more diffuse and decentralized. Video games are the most obvious example here, offering a main story quest that is complemented by side quests and non-player characters in order to create the simulacra of a living world. Today, those side quests are written by a team of writers under the direction of a creative director, and in the best cases, will carefully integrate with the rest of the game.
But AI will soon allow for a much more personalized gaming experience. Maybe instead of the main story, a player is intrigued by a neighborhood in one of the game’s cities and just decides to settle down there. For that player, a role-playing action game like Final Fantasy has just become a healing game like Animal Crossing, a genre known as iyashikei in Japanese. Or perhaps the flexibility in design afforded by AI could allow the player to select the genre they want to enjoy each day they play. In all cases, the vision of the creative director has been expanded, and they are no longer in control.
With Riskgaming, AI offers many of the same opportunities. Decisions in early scenes could completely rewrite later plot lines. Instead of a handful of available options for each decision, AI could be open-ended, calculating the possible response to the game’s core metrics. Decisions that I haven’t offered could be proposed, and the game could theoretically accommodate them. Players could spend resources like money in much more novel and interesting ways.
In short, we will soon see the rise of autonomous art, media that adapt and evolve to the audience under some strategic direction from the artist, or perhaps not even that. What was once my dictatorial laboratory ushering test subjects from place to place will soon become an open sandbox for player experimentation.
The compromise, though, is that the narrative imbued in art will fade in sharpness, and it will be different from person to person. Everyone viewing Guernica in the Museo Reina Sofía today may have their own response to Picasso’s masterpiece, but everyone at least saw the same canvas. Autonomous art means that the game, show, movie or book an observer enjoys is not the same work seen by anyone else. Your experience is your own. The disintegration of a common canon that I noted back in “Atomistic Literacy” will only accelerate further.
Many artists revile this trend. Some want to fight it tooth and nail. My own hesitation has given way to a sardonic acceptance of the reality that’s coming. I do think some of us will miss the purity of art that is created by a single author without assistance, and we will realize that a certain qualitative value of creation has been lost forever. As with all media though, the passage of time changes what the audience wants, and artists ultimately want their work to be noticed. At least the questions about AI will stop.
Podcast: Why immersion — and not realism — is critical for wargaming
Laurence Pevsner and I are still on vacation, so once again, we bring back our independent Riskgaming designer Ian Curtiss as guest host, this time to interview David Banks. David is senior lecturer in wargaming at the Department of War Studies at King’s College London, where his research focuses on the empirical evaluation of war games and how the craft can evolve in the years ahead. He is also the academic director of the King’s Wargaming Network.
🔊 Listen to “Why immersion — and not realism — is critical for wargaming”
Here’s a condensed and edited extract of their conversation.
Ian Curtiss: Two things I'm curious to touch on:
One is the sense that a wargame has to be a serious thing. One subconscious assumption might be that if you're having fun, then you can't be learning.
It is also so interesting that this methodology is so frequently used in war, in the defense industry, and yet this methodology is still frequently pushed aside by other industries because it can't be a serious thing.
David Banks: I think you're right, your point about fun. Right now, the term I tend to use in an effort to make it sound a little bit more expansive or scientific is “engagement.” Because you can also be playing a game that you're really into, but you actually may not be having a fun time. You might actually be getting quite stressed or worn out or exhausted, but it doesn't mean you don't still want to participate.
Engagement pulls people in, but I think on one level, embarrasses people because then they go, "Well, this is childish. This is for kids."
But instead of hiding from the fact that games are engaging, I think we should just say, "Yeah, that's right." In my estimation, that's one of the method’s key distinguishing factors. Not only does it engage, I would contend that it must engage to produce educational or analytic outcomes.
But back to your military bit, which is why does the military like this so much if at the same time it has a mild allergic reaction to it? I think one of the reasons is because gaming is good at modeling kinetic things in a consistent way, in a way that we have a lot of validity about the results.
Take a video game like Microsoft Flight Simulator. You can learn how to fly by playing that game. We've got such confidence about the internal validity of the models we use for that realm that we can actually build a really effective facsimile.
Where gaming starts to have much more questionable applications is once you leave the kinetic realm behind, because we don't have much confidence about other factors, like how do I know I'm looking at a democracy? How do I know I'm looking at the national will? How do I know I've identified interest groups correctly and the way they work together?
Ian Curtiss: This gets to a second point that I was going to raise, to your comment about immersion. It is one of the cheapest ways to get people to actually think about a future scenario.
To spend a solid two hours thinking about a potential future scenario that forces you to wrestle with the pain, the cortisol, the stress, the dopamine. So in the end the player is already asking, well, would it work that way? Would it not? And so the value of gaming is worth more than killing oneself over building the perfect model.
David Banks: Yeah. We can talk about building the perfect model. Firstly, we don't have a consistent evaluative metric to identify how I would know my model is perfect. My North Star is always asking, "What's the purpose of my game?"
You should be able to articulate that purpose reasonably clearly and then that can help you know, "Hang on a second. Am I going off-piece?" Am I adding design features to a game that really don't serve any function? Why am I doing that? We might be doing that for immersive reasons, but in terms of the game mechanics, if it isn’t core to what I want to study, I'm just adding complexity.
To your point about immersion, I think the argument is that immersion is what makes gaming a better method than alternatives under certain conditions. I'm paraphrasing Thomas Schelling, the economist, and I'm getting the quote slightly wrong, but the one thing that nobody can ever imagine is something that never occurred to them.
That's what gaming gives you. You can play a game of chess against yourself in terms of testing out strategies, but you can't actually surprise yourself. But when you're playing against somebody else, they might play better than you expected or worse than you expected, but they won't play exactly like you expected.
The OB: Is data-driven VC dead?
This week, Lux’s scientist-in-residence Sam Arbesman interviewed Lawrence Lundy-Bryan, a partner at Lunar Ventures, on how OpenAI’s Deep Research product will transform the future of venture capital. We last talked to Lawrence in August 2023 on “We need to go deeper with the inception of deep geothermal energy.”
🔊 Listen to “Lawrence Lundy-Bryan on how data-driven VC is over” over on The Orthogonal Bet.
Lux Recommends
- Peter Hébert noted the changing patterns of work in a recent survey by ActivTrak. “ActivTrak’s report found that while weekday hours have declined, weekend work is on the rise. Mauch said there will always be a core group of people who work on the weekend, and overall, people have grown more comfortable with flexibility, so many don’t mind leaving work early on weekdays and leaving the remaining work for weekends.”
- Our editor Katie Salam recommends a fascinating essay from John Last in Noema on “What Feral Children Can Teach Us About AI.” “This leads him to some radical and uncomfortable conclusions. Pleasure and pain, he argues, are dependent on the existence of this conscious, thinking self, a self that cannot be observed in young infants and animals. Does that mean Genie and Victor did not suffer from their abandonment just because they appeared incapable of performing mental synthesis?”
- Laurence was deeply moved by this profile of South Korea by Gideon Lewis-Kraus in The New Yorker on "The End of Children.” “Four out of five children in Korea today describe school as a ‘battlefield.’ In 2012, the advocacy group World Without Worries About Private Education helped develop an ad campaign that showed a baby bottle full of fried rice, with a caption that read, ‘Mom! It’s too early for me.’ Curfew laws prohibit hagwon classes after about 10 or 11 p.m. The issue nonetheless remains a society-wide prisoner’s dilemma, and even those who strenuously object in principle frequently relent in practice.”
- In a similar vein, Katie points to Ruxandra Teslo’s analysis in the latest issue of Works in Progress about “Fertility on demand.” “But the gap does hurt women in ‘greedy careers’ – careers that are greedy for the employee’s time and pay more per hour to the people who work the most. Consider a corporate lawyer working on a deal. The initial hours are spent getting familiar with the material and the people involved. Later hours – once the lawyer understands the case – are much more productive than those at the start. A person working a 40-hour week in this scenario does more than twice as much work as a person working a 20-hour week. Lots of careers are like this.”
- Finally, Tracie Rotter recommends Short Wave’s podcast episode on “Stone Age to Bone Age?.” “… a new discovery out this week in Nature suggests early humans in eastern Africa were also using animal bones – one million years earlier than researchers previously thought. The finding suggests that these early humans were intentionally shaping animal materials – like elephant and hippopotamus bones – to make tools and that it could indicate advancements in early human cognition.”
That’s it, folks. Have questions, comments, or ideas? This newsletter is sent from my email, so you can just click reply.
