英伟达公布AI黑科技！可生成超逼真人脸，还能神速修复受损照片

Can YOU spot the real person? Creepy AI can now create ‘100 per cent lifelike’ human faces from scratch (as well as animals, cars and even bedrooms)

你能分辨出哪个是真人吗？令人毛骨悚然的人工智能现在可以从零开始创造出“100%逼真”的人脸（还有动物、汽车甚至卧室）

素材来源：每日邮报 翻译：世界播

Can you tell who is real and who is not? Artificial Intelligence is now able to create lifelike human faces from scratch.

你能分辨出谁是真实的，谁不是真实的吗？人工智能现在能够从头开始创造出逼真的人脸。

Researchers at NVIDIA have been working on creating realistic looking human faces from only a few source photos for years.

多年来，英伟达公司的研究人员一直致力于从几张源图片中创造出逼真的人脸。

For many people it's difficult to tell the difference between one of the faces generated below and an actual human face, can you spot which is which?

对于许多人来说，很难区分下面生成的一张脸和真实的人脸，你能分辨出哪个是哪个吗？

The team at NVIDIA, released a paper on the subject, and explained they used Generative Adversarial Networks (GAN), to customise the realistic looking faces.

英伟达团队发表了一篇关于这个主题的论文，并解释说他们使用了生成对抗网络来定制逼真的面孔。

The fake faces can be easily customised by using a method known as 'style transfer' which blends the characteristics of one image with another.

通过使用一种称为“风格转换”的方法，可以很容易地定制这些假面孔，这种方法将一张图像的特征与另一张图像融合在一起。

The generator thinks of the image as a collection of three styles, known as coarse styles (pose, hair, face shape), middle styles (facial features and eyes) and fine styles (colour scheme).

生成器将图像视为三种风格的集合，即粗糙风格（姿势、头发、脸型）、中等风格（面部特征和眼睛）和精细风格（配色方案）。

Animals, such as cats, and objects such as a bedroom can also be generated, using the same method.

也可以使用相同的方法生成动物（如猫）和物体（如卧室）。

The researchers created a grid to show the extent to which they could alter people's facial characteristics using only one source image.

研究人员创建了一个网格来显示他们仅使用一幅源图像就可以改变人们面部特征的程度。

One of the most fascinating aspects of this is GAN has only be around for four years.

最吸引人的一点是生成对抗网络方法是4年前才提出来的。

But, is it not yet perfect, there are giveaways that can indicate that you are looking at an AI image.

它还不够完美，有露馅的地方，暗示你正在看的是一个AI图像。

For example the hair is very difficult to replicate, and as such, can often looked painted on, or slightly peculiar.

例如，头发是很难复制的，以及其他像这样的特征，通常能看出来是被画上去的，或略有古怪。

The advances in this technology also poses interesting ethical questions.

这项技术的进步也引发了有趣的伦理问题。

Can people really trust pictorial evidence?

图片证据真的能相信吗？

What are the implications for governments or repressive regimes being able to use this technology for propaganda or to spread misinformation?

政府或压制性政权能够利用这种技术进行宣传或传播假信息，这意味着什么？

Earlier this year we revealed how Nvidia software uses AI and deep-learning algorithms to predict what a missing portion of a picture should look like and recreate it with incredible accuracy.

今年早些时候，我们揭示了英伟达软件公司是如何使用人工智能和深度学习算法来预测一张图片缺失的部分应该是什么样子的，并以惊人的准确性重建了它。

All users need to do is click and drag over the area to be filled in and the image is instantly updated.

所有用户需要做的就是点击并拖动需要填充的区域，图像就会立即更新。

As well as restoring old physical photos that have been damaged, the technique could also be used to fix corrupted pixels or bad edits made to digital files.

除了修复损坏的旧照片，这项技术还可以用来修复损坏的像素或对数字文件的错误编辑。

Photoshop could become a thing of the past as new technology has been developed by Nvidia which can instantly improve touch-up damaged photos in seconds

英伟达开发出一种新技术，可以在几秒钟内迅速修复受损照片，Photoshop可能会成为过去

Graphics specialist Nvidia, based in Santa Clara, California trained its neural network using a variety of irregular shaped holes in images.

英伟达公司位于加州圣克拉拉，是图形处理方面的专家，它使用图像中各种形状不规则的小孔对其神经网络进行了训练。

The system then determined what was missing from each and filled in the gaps.

然后，系统确定每一项都缺少什么，并填补空白。