Are we headed towards a new “AI winter”?
AI NPCs.AI scriptwriting.AI voice acting.AI artwork.AI creation tools.
Yes, AI is everywhere in game development.
In the past couple of years it’s arisen as something of a dirty word, an inevitable future.
But the future is already here.
What does an AI future look like?
And why, now, is the progress of AI for positive means also under threat?
To find out more, I spoke with AI expert Dr. Tommy Thompson at the London Developer Conference.
So-called deep learning, meanwhile, refers to the additional complexity gained via artificial neural networks.
Generative AI is a subset of deep learning that adds contextual relevance.
An example of this is a language model like ChatGPT.
Generative AI can be used as a shortcut, then, but the results are often lacklustre.
Some of it may have been stolen and scraped, and should not legally be used.
In-line with Unity’s report, animation is a key area - particularly animation blending.
Games like The Last of Us Part 2 andHitman Absolutionuse this process, among others.
Using these systems allows developers to spot issues at scale.
Crucially, humans are still required as part of these processes.
A further example of this is SpeedTree, a generative tool for building trees in games.
“How much money do we need to throw behind this before we start actually getting returns?”
“And that’s a justifiable question.
It seems like for a lot of these companies it’s not quite there yet.
Crucially, the groundwork is already being laid for better regulation - be it EU laws or Steam approvals.
But how can they trust that a third-party tool won’t take that data to train their own system?
The law is a key contributing factor, with tools needing to comply with EU copyright laws and regulations.
Transparency of datasets and processes is needed, which third-party tools cannot always guarantee.
Another question: how can companies guarantee these third-party tools will still exist in a couple of years time?
And that’s before staff costs too.
Earlier this year,Stability laid off 10 percent of its stafffollowing the exit of its CEO.
And when that falls through, it leads to an AI winter.
But for indies, perhaps the biggest hurdle isValve’s approvals process for Steam.
This identifies two types of generative AI: pre-generated and live-generated.
What’s the future for AI in games?
Thompson agrees the results of experiments like this have been poor.
“I think that’s an equally valid question.
You couldn’t even bother to take five minutes to write it.
Would an AI really capture what makes these different franchises great in a mash-up of the two?
Would it understand the tactile controls, the unique visual aesthetic, the kinetic and frantic combat?
“I don’t think [AI] would inherently understand the quality of those,” says Thompson.
“There’s a lot of pessimism, but also increasing apathy towards generative AI,” says Thompson.
“We’re in such a hype cycle right now.”
It certainly feels that way.
“Advancements in AI will create more personalised experiences and meaningful stories for consumers,” Qizilbash said.
It sounds flashy enough, but can companies really deliver on these buzzwords and statements?
Or, to paraphrase Game of Thrones, is winter coming?
“The proof,” says Thompson, “will be in the pudding.”
