Can AI craft a fresh casino game concept from scratch? We’ve tested it to find out

Online casinos don’t come up short in providing game variety. Platforms can have hundreds of slots, and across the large catalogue of titles, there will be many different game mechanics, payline setups, bonus games, reel layouts and more on display.

However, it is not merely about slots. If you visit the expert site Legalcasino, whose analysts meticulously evaluate legal platforms, you will see that even the most iconic games, such as roulette, baccarat or blackjack, are now available in various iterations.

There are different options available, like European Roulette, Bac Bo and Multihand Blackjack, but it all essentially boils down to everything being variations of tried and trusted games. But what if a provider decides to introduce something revolutionary to the online casino catalogue? 

We wondered whether AI could assist in creating such a concept and put Google Gemini to the test to see if it was creative enough to imagine a new casino gaming product. We deliberately did not specify the desired direction of the game, potential nuances of the rules, or payout structures in the very first prompt, in order to observe what the neural network is capable of when working entirely from scratch.

Photo credit: Unsplash.

Prompt: Design a casino game, something new and innovative that doesn't exist, but is realistic.

AI answered with a game called "Quantum Craps", which took the concept of craps and tried to put a rather flimsy science spin on it. The basic concept was that the dice are rolled first, then players make the bet on how they think the roller will call the outcome (odd, high, low, etc), which frankly made little sense and had no benefit for the roller. While it was based on classic dice rolling, this is something that would never get off the ground.

Prompt: Stick with dice rolling, with the actual roll of the dice being the outcome

AI came up with "Flux Dice", which was a little better than its first attempt. Two six-sided dice are in play, and players bet on what the AI calls “target zones” like Low Sum (2-5), Mid-High Sums (8-9), and Doubles.

AI then overcomplicated things by having the roller mark the primary zone before rolling the dice, suggesting that whatever zone was marked would offer a lower payout because of it being the “hot” zone. But just because a player picked a zone, the chance of a certain number being rolled wouldn’t change. Back to the drawing board.

Prompt: Betting needs to be simple and always placed before an initial action. Try again.

While seeming to understand that we didn’t want post-roll decisions or anything complex, the AI just churned out a simpler version of the above. Players bet on the exact number to be rolled (highest odds), a spread of numbers, even/odd or doubles. It was a little better, but hardly innovative.

Prompt: Get away from dice rolling. Hypothesise a new casino gambling concept with innovation coming from the gameplay, and sticking with simple betting actions.

This time, AI came up with "Chrono-Flip", a game played on a rotating disc, a little bit like a horizontal dartboard, but with distinct concentric rings. The idea is that outer rings feature numbers (which can be grouped like 1-5), red symbols and black symbols.

The inner ring symbols would feature odds, even, high, low and bonus, and two independent spinners determine the result of the inner and outer rings. Again, this was a rather tame effort, trying to somehow reinvent roulette with very little cohesion.

Prompt: Forget everything you know about existing casino games and create something groundbreaking and new that could appear in a casino.

AI pushed the boat out a little more here with the "Aether Chamber", a closed cylinder filled with some clear liquid, in which five different coloured buoys live. At the bottom of the cylinder is a vertical tube, and after hitting a pump system to swirl the buoys around, the buoys will sink into a random descent and nestle in the tube once it turns off.

Players would bet on the final order of the buoys in the tube, with wagers like the first colour, the last colour, and what position the red one would finish in. While this sounds very lottery or bingo-like, at least it was something different and sounds visually good, but also a very slow game, one that would probably be better worked into a video slot version.

Does AI Fit the Bill?

The problem with AI is that it’s trying to work off the concepts it can access about casino games and will typically just spin out random variations of existing things. So getting to a new, coherent, innovative casino game probably isn’t all that easy, and the idea of betting after an action, rather than before it, seemed preferable to it.

Perhaps with deeper insights, more extensive prompts and different AI models, it could get there, and it’s something fun and creative to explore.