Microsoft I presented its last project in the field of artificial intelligence – โThe Model of the World and Humanโ (World and Human Action Model (Wham), which can generate new gameplay personnel based on videos of real players. Despite the fact that the technology is developing rapidly, it is still in the early stages, so do not pay attention to the โshakalโ quality.
Analyzing the actions of gamers with videos, Wham โrecognizesโ how gaming objects and characters behave, which allows it to generate new scenes. To do this, it is enough to send the models a screenshot on the basis of which personnel will be created. This process can even be controlled, for example, by a gamepad.
Research Microsoft showed that WHAM can maintain the โconstancyโ of the gameplay process, that is, to have a logic of actions without leaving in โhallucinationsโ, within two minutes.
One of the most important features of WHAM is the ability to โmemorizeโ and use new objects introduced into the game. It is as simple as copying and inserting the picture.
However, the technology is not yet ready for real -time use. The current prototype works with a low resolution of 300 ร 180 and a frequency of 10 frames per second.
February 20 The Thing will receive a physical publication on June 25
February 20 of the media: Netease loses interest in foreign developers
Bleeding Edge
The best comments
A new generation of bots in a multiplayer is coming.
And billions of people on the planet will have a friend Victor, who, unlike all these meat bags, understands you well and always on your side.)
However, the technology is not yet Grand-Hotel-Casino.co.uk ready for real -time use. The current prototype works with a low resolution of 300 ร 180 and a frequency of 10 frames per second.
In the sense of "not ready"? Apskail + Dlss + Personnel Generation Threw and Refine.
It remains to integrate a chat both there with a simulation of a voice chat with a roots, and, consider, a cure for the syndrome "no one to play" was found with
It is surprised that the numbers on the screen are constant, that is, the CP decreases from damage and the kuldowns are rolled in the right direction.
And then similar to the oasis of minecraft is broken tightly and the interface is unstable there, not counting the whole environment at all.
You can try the link, I advise you to periodically look at the feet. Main challenge – build a tower under you.
Mines and dig dynamite))
I don’t quite understand why this is necessary – neural network. making the vidos of gameplay any game. Teach her to generate a completely new game? Generate in real time so that it can be played in it? It is difficult to understand.
She already showed this in Crackdown 3.
Somewhere there was progress ..
From the point of view of checking the power of neural networks – yes. It’s good … but dead end. Conceptually it is like exploding dynamite to dig mine. Yes, it works, but it is better not worth.
It is instead of inventing models, code code, to build a card (and the neurons can do this, I believe in it) just draws shots of the video based on the previous ones, scoring what was before.
There was already some kind of rumor that they want to use the neural network so that in online games there were companions with AI
Stop at the stage of frame generation, please. And so the FPS on the counter is already fake, so you are now also gameplay, and the fake physics intends to saw.
So the game is not generated, they teach AI to repeat in a real game, which, in fact, is an overshotch clone
I think, then they will teach others, from something you need to start
It was still a neurodum
If this helps to make good games faster, then why and yes ๐
They have in their hands the potentially unlimited capabilities of AI and all that they can generate is another overtubid clone?
T -shirts themselves say that this can be used for ports of old games for new devices
It looks like a feces
I did not understand from the article what is the task of this neuron. If this is an analogue of Dlss, then why should she generate two minutes of gameplay for two minutes. And if it is used to generate a video in screenshots, then it is difficult to figure out why this is needed. Can throw ideas, as if in 100% of cases they will receive Generick Hernya.
