AI nears a tipping point for games developers

Artificial intelligence (AI) may not be at the world domination stage but nonetheless its reach has expanded hugely in the past year. Programmes that scan hundreds of millions of images online to turn text prompts into new pictures are now widely available online. One industry this is already impacting is gaming: video game developers are already using the tools to inspire creative artwork. They expect AI to produce further productivity gains as the technology develops. 

In April, image generator DALL-E 2 was released by OpenAI. It was a big jump forward from DALL-E, which was released in 2021, as it allows the merging of two images. For example, rather than just generating a picture of a monkey, it can now generate one of a monkey on skis in the style of Van Gogh. As pointless as that sounds, it has had a wider impact. DALL-E 2 is trained using hundreds of millions of images on the internet and a new image generation process called ‘diffusion’. This may not seem groundbreaking but it was a revelation when released – the spooky nature of the images has also raised questions about what can be considered art. 

Two similar services, named Midjourney and Stable Diffusion, which both have similar capabilities, followed in July and August. For people in the industry this year is seen as a watershed moment. “The work of DALL-E 2 and Midjourney is actually game changing, they have capabilities we have never seen before and appeared seemingly overnight,” said Danny Lange, senior vice president of AI and machine learning at Unity Software (US:U), a $9bn (£7.8bn) market cap video game software development company. 

In an industry where production costs are extremely high, artists are already using these programmes to speed up their work and help with creative inspiration. As the programmes can only produce 2D images they are currently most useful for flat backgrounds. “All our artists are using programmes, such as Midjourney, particularly for mood art,” said TinyBuild (TBLD) chief executive Alex Nichiporchik. “But they are also useful for sparking new creative ideas”.

Gaming could be a big beneficiary of AI because development costs are so high. Blockbuster titles, know as AAA, can cost up to $80mn to create. If a big game flops, developers can immediately swing from a profit to a loss. Last year, Frontier Developments (FDEV) saw its operating profit fall from £19.9mn to £1.5mn because of the failure of its Elite Dangerous: Odyssey Game. It ended up having to take a £7.4mn impairment. Reducing production costs would shrink the potential downside if a game were to flop like this. 

At the same time, artists are concerned they could soon be replaced. However, the tools are currently more an aid than a replacement. Writing interesting prompts is a creative skill and, in the future, Lange thinks “prompt writing” will become a job in itself. “The output of the programme is only as good as the input of the artist,” he said. 

Big steps to come

Video games are 3D animations so there is a limit to how much existing image generators Midjourney and DALL-E 2 can speed up the development process. Most of the time and money spent in development is on improving animations. Nichiporchik believes the major steps forward will occur when AI can create 3D images and animate the characters. “When AI can be used to create a character that moves organically, it will take a lot of the procedural work out.”

It is not certain when this breakthrough will happen, but once one discovery is made, other companies can build on that model. Code is often published online for this reason. OpenAI built DALL-E 2 based on the diffusion image synthesis model created by Google Research in 2015. There are already papers published outlining methods for generating 3D imagery; however, the process is currently limited to low-resolution images because of the computer power required.

But this cross-sector development is the way forward: innovations made with driverless vehicles could be reapplied in the gaming industry. Software that can help cars more through a real environment is also useful for helping games characters move through a digital world. “It will only take one leap from the big boys like Google or Tesla; they will then publish that research and then it will be used through the gaming industry,” said Nichiporchik.

In this situation, the characters and environment within a game could be generated in real time as the gamers play. TinyBuild has already published a game called Streets of Rogue which uses AI to generate levels for the players. The game was created by just one developer, Matt Dabrowski. 

Streets of Rogue is a top down 2D game, so the processing power needed to play is relatively low. However, current computing power would run out of puff if games were 3D and fully generated by AI as the player moved through them. Both DALL-E 2 and Midjourney need to be run on cloud computing networks because localised processing is not powerful enough. This wouldn't work with real-time gaming because time needed to send the data two and from the cloud would create a lag.   

More power

Increased localised computer power is necessary and currently the ability to make microprocessors any smaller is hitting a physical limit – Moore's Law, which forecast the doubling of the number transistors that fit on a microprocessor every 24 months, is coming to an end almst 60 years after it was first identified. The big chip designers such as Qualcomm (US:QCOM) and Nvidia (US:NVDA) know this is a chance to differentiate themselves from the competition. “All the hardware manufacturers are aware of the need to optimise for AI and mobile devices are now reserving space on their chips for real-time machine learning,” said Lange.

The need for localised computing is also a future problem for Microsoft’s (US:MSFT) Xbox Game Pass. The pass allows customers to access games through the cloud – as opposed to downloading an entire game or playing it off a disc using a console – and can even be played over new-model Samsung TVs. The upside is that only a controller is needed, so hardware costs are low for users. However, it may find itself limited to more basic games, if the developments in software require ever-increasing amounts of computing power.

Lange thinks that cloud computing will be problematic for high-tech games, but also warned engineers in the space should not be underestimated. “You could theoretically use AI to predict where the gamer will walk, so it could generate their environment before they move which would remove the latency issue between the cloud and the device,” he said. 

Frontier Development’s chair and founder, David Braben, is less euphoric, positing that the recent progress in AI is not a watershed moment, but just another tool for developers to use. “Our industry is one of the fastest developing and every year there is an innovation, AI is just placing more ability in the hands of talented people,” he said. 

Whether 2022 is viewed as a pivotal moment in the life of AI, it is clear it will be useful in the future development of games. Improved AI will lower costs, but this will happen across the industry, meaning margins could be competed away on price. 

The main returns will likely accumulate further up the supply chain – with the hardware company that produces the equipment that can support AI innovations. Nvidia is well placed as the market leader in both graphics processing units and AI chips. Last month, it released a new chip that uses artificial intelligence to enhance graphics. The graphics cards known as the ‘Ada Lovelace’ will be produced by Taiwan Semiconductor Manufacturing (US:TSM).

"Games developers are famous for pushing hardware to the brink and AI is adding a new dimension for them to do that," said Lange. It will be fun to see how far they can go.