THOUGHT

Thank you, gamers

70 years of AI research. The real heroes of the story never read a single paper, but they had excellent reflexes and a strong opinion on frame rates.

A light ink drawing of a relay race on a sunny track: a teenage gamer hands a graphics card, used as the baton, to a running scientist in a lab coat, while gamers in the stands cheer with controllers raised.

Over 70 years of AI research, the real difference has come from people who played Doom or, more broadly, games that required complex 3D graphics, like Quake III Arena, Doom 3, or Crysis.

The presence of the gaming market actually helped develop two kinds of resources for what we have today:

For one, and this is a bit geeky, the keyboard. We wouldn’t have mechanical keyboards anymore if not for gamers wanting minimal latency. People like me prefer mechanical keyboards because they feel much better than soft membrane keyboards. To me, those keyboards are too slow and inaccurate, while mechanical keyboards are much better. They’re loud, yes, but that’s part of the appeal.

Mechanical keyboards were practically extinct at a certain point, but competitive gaming centered on having minimal input latency. Keyboards with mechanical switches became a viable market that justified new companies and startups building them, and now we have a lot of choices.

The second important point is about AI. Today we take for granted that industry leaders like AMD and especially Nvidia produce GPUs that are exceptionally well suited for artificial intelligence. But at the beginning, a GPU was simply a graphics processing unit. It was built specifically for video game graphics: Sony coined the term GPU in 1994, for the graphics chip in the first PlayStation. The gaming market allowed Nvidia to grow and secure the position it holds today: I’m not sure they would have survived without that demand, which continuously pushed the need for faster GPUs. Now Nvidia and AMD hold leadership positions in producing the cards that are essential for AI.

It is ironic that the prior existence of these cards made the take of of AI possible, enabling the deployment of massive computational power across existing machines in a way that would have been impossible otherwise. The same goes for the adoption of CUDA (the Nvidia platform to use its GPUs for general computing) and the software tools that standardized how AI engines operate. Without large players driving this ecosystem, we would not have the mass market today, which also leveraged the standardization on a common platform and the availability of advanced hardware at an affordable price.

We probably would have seen a very different story. The algorithms invented in recent years might not have been adopted at this scale, simply because speed is critical. If an AI system in the year 2000 took an hour to produce a single paragraph, we might have thought it was wonderful, but we would never have found a practical use for it. Today, generating thousands of lines of code in hours is a game changer.

That is the main thought for today: after 70 years of AI research, we owe a thank-you to the people who simply played video games.

Thank you, gamers!