AI-First Game Development. Part 4. The Ideas Machine
Welcome to Part Four of my series, AI-First Game Development where I try answer the following questions:
Can AI help a rusty ex-coder with a passion for games, not a lot of free time with work and family, create a decent computer game? And if so, how much load did the AI take, how has the software engineering process changed since I was in the driver’s seat, and what other aspects of game design and development can AI help out with beyond writing code? And finally, what are the best tools out there to help me with this job?
If you want to catch up or skip ahead, here is a list of all posts in the series.
Part 0: AI First Game Development (Intro)
Part 1: AI-First Game Development. Training the GPT
Part 2: AI-First Game Development. My AI Teacher
Part 3: AI-First Game Development. The Chasm Between Design and AI-Generated Art
Part 4: AI-First Game Development. The Ideas Machine
Part 5: AI-First Game Development. The Engineer Lives
The AI Muse
Creativity along with our opposable thumbs are attributes that set us apart from all other species. Not only have we produced breathtaking art, music, and literature, but also mastered game console controllers making humankind the undisputed champions of Fortnight.
I’ve always thought of game design and development as the pinnacle of creativity in the digital age, especially in the early days of computing where the gap between creative ideas and technology capability was vast, and true innovation was required to squeeze joyful experiences out pitiful amounts of memory and CPU, and very crude software tools.
Today it’s different as we have a glut of compute resources in the palm of our hand, and game engines like Unity and Unreal Engine to help us create incredible graphics and game play with relative ease.
We also have a new tool in our box to help us with the creativity process.
Hello again, Generative AI.
For those who haven’t been following the series, I have been using Gen AI to help me build a mobile game, Bubble Pop Hero, and am approaching this in an AI-first way.
Essentially this means asking one question with each task, how can AI help? And then experimenting.
For tasks that require creativity like coming up with game concepts, game mechanics, names, and level ideas, Large Language Models (LLMs) are a powerful tool to generate ideas from basic prompts, turning ChatGPT and the like into your AI muse.
Here are some examples of how AI has helped me with the creative process in designing Bubble Pop Hero!
Game Conception
In an earlier post, I mentioned that Bubble Wrap Hero was an idea I had long ago, so I can’t claim AI helped me there (although perhaps it should have) .
But, since my GameSpark GPT has been tuned to be a Unity expert experienced with building casual games with examples, it has been indispensable when it came to fleshing out of the ideas. (See sections below. )
But, were I to create another game in the future I would use GameSpark GPT to suggest different ideas to see if there is something better to invest my time in.
It’s also a great opportunity to tweak some of the settings like Temperature to dial up or dial down the creativity of the LLM response.
Here are responses with the low and high temperature settings for the following prompt. You be the judge which is better. Not that much in it, IMO.
Prompt:
give me 5 fresh ideas for a fun and addictive hyper-casual mobile game.
Standard Temperature (0.1) (Less Creative)
Sky Glide: Navigate a glider through rings in the sky, avoiding obstacles.
Infinite Tunnel: Dodge obstacles in a spinning tunnel, speeding up over time.
Color Swap Dash: Match character color to walls to pass through them.
Zen Slice: Slice through objects with precision for points.
Bounce Master: Guide a bouncing ball from platform to platform, collecting coins.Higher Temperature (0.8) (More Creative)
Skyline Skater: Skate on a changing skyline, avoiding obstacles and collecting coins with simple swipe controls.
Color Splash Dash: Navigate a ball through color gates in a 3D tunnel, matching the ball's color to pass through gates.
Zen Garden: Maintain a virtual zen garden, raking sand and placing elements, with challenges like leaves falling to disturb the peace.
Infinite Racer: Drive on an infinite road that changes width with speed, dodging obstacles and collecting power-ups.
Bubble Ascend: Guide a bubble through obstacles by tapping to rise or fall, with power-ups to aid progression.Game Mechanics
I wanted to make sure that Bubble Pop Hero was fun and addictive. To provide guidance, asked GameSpark GPT the following prompt and it provided some good general suggestions.
Prompt:
I have the following game but need help with making it more sticky / addictive. in the game, you have a grid of bubbles that you tap to pop with your finger. behind the bubble can be a collectable, a bomb, a clock, or empty. there is a timer bar at the top of the screen which runs down. the timer starts somewhere in the middle. if the timer gets to 0, you lose the level. if you add enough time and max the timer out you win the level. if you hit a bomb, the timer counts down faster.
Response:
- Progression and Levels: Introduce levels with increasing difficulty and new challenges.
- Rewards and Collectibles: Offer collectibles to motivate ongoing play.
- Power-ups and Abilities: Provide power-ups for enhanced gameplay.
- Daily/Weekly Challenges: Feature challenges for unique rewards.
- Leaderboards and Social Features: Add leaderboards and social engagement features.
- Customization Options: Allow customization of game elements.
- Mini-Games or Side Quests: Introduce alternate gameplay modes.
- In-Game Events: Organize special events for new content.
- Feedback and Progress Indicators: Provide feedback and progress tracking.
- Balancing and Pacing: Ensure the game is well-balanced and paced.
- Tutorial and Help Features: Offer tutorials and help for new players.
- In-App Purchases and Ads: Consider monetization through purchases or ads.In-App Purchases and Ads: Consider monetization through purchases or ads.Fairly broad, but I could then drill into each area with a lot more back-and-forth.
Coming Up With Names
There was a small naming crisis where Daz thought Bubble Pop Hero wasn’t the best name, and convinced me we should think of something better. The reason is there is a ‘Bubble Pop’ style of game that is almost a genre. We didn’t want to get lost in the noise.
Long story short, GameSpark did come up with many other suggestions but we opted to stick with Bubble Pop Hero!
Another area where GameSpark helped was with creating level categories. I was looking at the 90 odd animals we had (which would be collected by popping the bubble, and thought animal types could be a good way of creating levels.
I pasted a screenshot of all the all the files into GameSpark and asked what classifications I could put these animals into.
GameSpark managed to read all the animals from the image and come up with a list of category suggestions. After a bit of back and fourth, I settled on six main categories and then worked with Daz to finalise which animals belonged where and what the category’s would be named. GameSpark did a great job helping with this and in a couple cases gave us a preferred suggestion.
Validation
Is an idea good? This has to be one of the most difficult questions to answer without developing it and testing it out in the world. I did ask GameSpark what it thought of my ideas, and provided mostly positive feedback and some useful alternatives.
Visual Inspiration
In the last post I talked about using the Image Generation capabilities of GameSpark, but found it lacking for creation product ready assets.
On the plus side, it gave me the ability to create proof of concepts that had a form for visual expression, not just place holders, that brought ideas to life more effectively.
Albert Einstein said “Creativity is Seeing what everyone has seen, and thinking what no one else has thought.” Applying this to LLMs, we know that the large language models have seen a lot of what everyone has seen (or read), but how good are they at thinking of what no one else has thought? This is a trillion dollar question and we will have to wait and see.
My feeling, we aren’t there yet, but it won’t be long.
Read on to part five, The Engineer Lives, where I look at how Gen AI has helped me with Engineering, and the humbling insights I gathered along the way.
