Javi Lopez’ AI-created game is one that elicits an immediate sense of recognition. It’s in the overall aesthetics, gameplay, and even the title: “Angry Pumpkins.” But despite it being so overly inspired by a very famous game (and franchise) that’s limited to throwing bird-like characters around, Angry Pumpkin’s true value lies in the way it was created, not with a team of programmers and developers, but with tools like ChatGPT and Midjourney for creating the code and graphics.
Using only natural language, Javi prompted ChatGPT 4.0 towards providing a complete, working 600 lines of code – the instructions to create an interactive version of “Angry Pumpkins.” Midjourney was used to create objects, characters, and backgrounds, while the Title Screen required DALL-E 3 from Open AI.
Games development is an expensive affair. Every year, we hear of gargantuan, multi-billion dollar acquisitions of game development studios – organizations that are frequently in the hundreds of team members. Conversely and unfortunately, we also hear of proportional layoffs.
But even beyond studios themselves, videogame budget allocations are massive. Star Citizen, the running marathoner here, already blew through an (inflation-adjusted) $541 million. Cyberpunk 2077 is estimated to have cost around $500 million to develop across consoles. Even the original Angry Birds – yes, that game released back in 2010 – cost $140,000 to produce.
Midjourney, DALL•E 3 and GPT-4 have opened a world of endless possibilities.I just coded “Angry Pumpkins 🎃” (any resemblance is purely coincidental 😂) using GPT-4 for all the coding and Midjourney / DALLE for the graphics.Here are the prompts and the process I followed: pic.twitter.com/st3OEhVVtKOctober 31, 2023
This happening is a necessity, as game development requires such a distributed, specialized environment, working across disciplines in what is likely the most cross-sectional representation of human creativity: video games. From writers to game-engine or systems coders, lighting and character artists through music producers, casting, and filming, and the logistics required to support all of that, modern video games require all disciplines of creativity and management to master and turn into a successful product.
And despite being so complex, games add something that cinema or music can’t: they add the element of interactability. Videogames are alternate realities you can act and feel within.
What Javi Lopez’ Angry Pumpkins shows is that AI has brought massive cost-savings to video game development. A single person can essentially create a game that cost $140,000 to produce in 2010 – all you need is access to a computer and an internet connection that ends on an AI chatbot’s prompt box. ChatGPT helps you along the path of creating the code. Dall-E or other image-generation techniques take care of the graphics. Other solutions – other AI agents – can help in other ways, such as mapping out development costs, producing music or realistic voiceovers, and, in the end, translating the entire game into a number of languages.
And you can do all that in the language you know – although it’s a bit easier and cheaper if your language is English, to be fair. Being able to interact and navigate these programming-only systems using our own natural language is the prize here.
A number of videogame mediums have become incredibly accessible as the variety and power of AI generation tools have increased. Visual novels only require writing prompts and screenplay; add a camera and an AI-based filter to any footage, and you can even create experiences like the latest from Remedy, the incredible Alan Wake 2. It’s all about prompt engineering – which means patience and iteration in guiding the AI, not fully trusting its answers, and giving it increasing layers of simple instructions towards achieving complex results.
First of all, would you like to play the game?Here’s a link! (Currently, it doesn’t work on mobile): https://t.co/tynYmxhLzMIf you read the text below the game screen, which provides explanations, you’ll see how you can create your own levels and play them! 🙂 pic.twitter.com/uB52GUI5FSOctober 31, 2023
We should always remember that cost-cutting does usually imply a loss. In the case of AI-generation-led game development, the cost is human artistry. As the cost of using AI-generation tools is already lesser and becomes even cheaper than hiring human specialists, pressures are introduced. And because one creator orchestrating multiple AI agents allows for a measure of full creative centralization, it’s likely that more and more people who become independent developers also choose to take the cheaper and more controllable road of AI-assisted development. It’s also likely that certain game development companies scale back in personnel as they, too, embrace these tools. It’s more of a matter of time than a matter of likelihood – and there’s a human cost involved.
More positively, this cost decrease should result in an increase in the number of game developers, of people creating games either for their own consumption or for sharing with others, and able to bring their own personal visions into a playable state.
It remains to be seen how exactly we’ll deal with AI-generated content and copyright infringement claims, but a recent litigation result doesn’t favor creators whose work was allegedly used to train an AI content generator. In it, Judge Orrick dismissed certain copyright infringement claims against Midjourney and OpenAI on the difficulty of establishing proof that the content was used to train AI, since that content isn’t included in the model itself. Other elements were at stake for the dismissal – but it’s clear there are still more questions than answers.
Game development, as we’ve known it, is dead. All hail game development!