At GDC in late March this year, AI was everywhere. There’s always a buzzword at GDC of course, with the previous years’ pandemic-hit shows being dominated by the opportunist grift of talks and stalls dedicated to web3, blockchain, and the metaverse. But this time, amongst the still lingering whiff of crypto-bullshit, artificial intelligence was the talk of the town. And this time it feels different. Where the blockchain chatter was very much a product of new, external parties stepping in to fill the space left by real developers during a quieter year, all the talk of AI was very much coming from inside the house.
Typically it would be worth drawing some context around the recent news in AI at this point, but the reality is that’s rather hard to do, precisely because of what AI is. If nothing else, AI as a topic is a rapidly moving target. What was once bleeding edge – remember Twitter’s brief dalliance with Dall-E? – already feels old hat. GDC’s context, though, was that of a week of developer talks around AI that took place within days of major announcements from AI’s biggest players. Microsoft unveiled its GPT-led Office companion, Copilot, on 16th March. Google announced something that sounds rather similar on the 14th. On 15th March, Midjourney, one of several viral AI image creators, introduced Version 5 of its tool, which is able to overcome some AI image generation’s earlier struggles, like celebrity faces and regular-looking hands.
For some illustration of the speed of change here: by 30th March Midjourney had paused signups for free trial users because of “abuse” of the system. Its creators explain this as being down to a surge of one-time users responding to a viral how-to in China, that combined with a temporary GPU shortage and caused it to crash – not, they emphasise, all the (fake) viral images of Donald Trump being arrested and various other celebrity deepfakes that started doing the rounds. (It also banned various prompts that might lead to “drama”, although the Verge noted there are easy workarounds here, such as the “Donald Trump being arrested” prompt being banned, but “identical output” being possible with “Donald Trump in handcuffs surrounded by police.”)
Just last week, the European Union proposed its first AI copyright regulations, which would require companies using generative AI programs, such as ChatGPT, to disclose any copyrighted material used to generate its output.
At GDC itself meanwhile: at least 20 talks dedicated to AI – that’s excluding the other AI, as in, the one describing non-human opponents you encounter in games that we also, confusingly, refer to as AI. Although there is some overlap there too. Plus, relatively major announcements from the likes of Unity about AI tools, and all sorts down on the physical show floor. At a glance, it means shows like GDC can present a fairly unified stance from the industry: AI is here! On closer inspection, we’ve seen that before – see Microsoft CEO Satya Nadella’s pronouncements of mixed reality’s arrival in 2019, the metaverse’s in 2021 and AI’s this January, for instance – and more importantly, it’s just not really what the industry thinks. Largely because, as one person put it, AI is not a “monolith” to be considered as one singular thing. But also because neither’s the games industry itself. The real reaction was as wide a mix as you could get. On the one hand: packed-out talks, glib company lines and a cultic kind of reverence. On the other: scepticism, disdain, and outright rage. And all manner of opinions in between.