Super Smash Bros. creator thinks AI will make a lot of games, most won't make money

[Alfonso Maruccia]

GenAI Doh: During his career, Masahiro Sakurai created a few extremely successful game franchises, including Kirby and Super Smash Bros. Now, the Japanese designer is weighing in on the generative AI debate, arguing that LLM-based "stochastic parrots" don't have what it takes to create a truly original game IP.

In a recent interview with Japanese website Denfaminicogamer, Sakurai shared his well-informed opinions on how generative AI is going to impact the video game business. Developing new game "products" is becoming easier than ever, he said, but turning AI-generated slop into a profitable venture will take a lot more than a few chatbot prompts.

Sakurai explained that modern game designs usually stem from direct comparison with what's already on the market. That gives developers and publishers an easy way to show customers what they're getting for their money.

In this regard, AI will make things much easier – and much worse for the gaming industry as a whole. Generative AI is set to trigger an explosion of lookalike games, Sakurai said, but most of these copycat "productions" will end up unprofitable.

The technology will certainly help mass-produce poor-quality content, with most AI games replicating gameplay mechanics players have already experienced elsewhere.

Sakurai is also concerned about the lack of truly creative projects in the mounting wave of GenAI gaming slop. Creating a genuinely original game is beyond what AI can do, he suggested, and extremely difficult for human designers as well. He cited Super Smash Bros., noting that the first game in the series had to take a truly novel approach to stand out from the traditional fighting game genre.

Sakurai's views on game design and creativity are bound to fuel an already heated debate. According to a recent CESA survey, more than 85% of Japanese developers now use some form of LLM-based solution in their workflows.

Epic Games, for its part, has argued that AI will be involved in nearly all future game production, while paying PC customers have increasingly lamented the impact of GenAI games on Steam's Next Fest event.

This year's Next Fest is suffering from the AI slop overflowing on Valve's digital platform, and users are now asking for an official "no-AI" filter to make interesting demos easier to discover.

Sakurai said Steam's numbers show how current development trends are inflating the market. In 2010, Steam hosted a few hundred new releases per year, around 275. Today, more than 19,000 new projects launch on the platform annually. Generative AI will only push those figures even higher.

Permalink to story:

Super Smash Bros. creator thinks AI will make a lot of games, most won't make money

 

[Inthenstus]

That’s not how AI works, you have the ideas, and AI helps YOU bring YOUR ideas to life. This guy.. fcking Nintendo.

 

[NumberNine]

Humans weren't making very many completely original video games before LLMs came along so I'd expect that to continue now as well. There are also countless games that were completely novel ideas and yet still missed the mark for most players, resulting in an experience that could have been fun, if not tainted by an outside influence - usually money.

LLMs can write code for you, but they can't fix poor taste in design. If an idea was generc trash before, vibe coding it won't fix that.

There's no substitute for human creativity, but there are plenty of force multipliers. Consider the amount of auto-tuned vocals used as a crutch for people who can't hold their own notes in modern music.

It will always come down to the proficiency and vision of the hand wielding the tool. How you do something, is everything.

It's also worth noting that Kirby is based on a character who copies the abilities of other characters and that most of the smash brothers roster is copied from other games.

 

[Frankenhole]

Its going to be just like what happened to music when home studios we cheap enough.

There will be so many crap games that you can't sort thru the pile to find the good ones.

 

[James321]

Its going to be just like what happened to music when home studios we cheap enough.

There will be so many crap games that you can't sort thru the pile to find the good ones.

Pretty much like that already. AI will just make it much worse.

 

[Melkor Unlimited]

LLMs can write code for you, but they can't fix poor taste in design. If an idea was generic trash before, vibe coding it won't fix that.

Human written code, vibe code, paid artist images, AI generated images, NCT, NR, DLSS, RR, FG, ... whatever else.
All of that and some may, or may not, (help to) create highly polished, boring, crap when the whole idea behind it is ... milking some money from customers.

 

[NumberNine]

Human written code, vibe code, paid artist images, AI generated images, NCT, NR, DLSS, RR, FG, ... whatever else.
All of that and some may, or may not, (help to) create highly polished, boring, crap when the whole idea behind it is ... milking some money from customers.

I couldn't possibly agree with you more. We're seeing remarkable things happening now, but even with the increased ability for more people to generate more software, the end result will be a direct reflection of the creator's standards and practices. For those coming in with little prior knowledge of the concepts required to fulfill their vision, they'd be served well by taking it slowly and learning as much as possible about the fundamental building blocks that make it possible.

For example, an artist vibe coding a game from scratch will likely have a game with exceptional art.
Likewise for a musician, gameplay developer, dungeon master, fiction writer, school teacher, etc.
The personality and skills will come through. If there are none, slop.

 

[Dreadcthulhu]

Its going to be just like what happened to music when home studios we cheap enough.

There will be so many crap games that you can't sort thru the pile to find the good ones.

Someone just needs to create an AI that can sort through the pile for you . . realistically, game stores like Steam or Epic should at least attempt to do something like that, for their recommendation systems. Those stores at least have a strong incentive to show you games you might actually want to buy.