这项AI技术能够描画出你脑海中想象的图像

关于人工智能的应用远远不止于做助听器和机器人等功能。人类一直想研究出一些可以读懂他们思想的技术。到目前为止已经进行了很多实验。就在最近，又一项实验出现了。

日本ATR计算神经科学实验室的科学家们创造了一个基于人工智能的系统，这个系统能够从人脑活动中进行深度图像重建。

用简单的话来说，它只是在“阅读”我们的想法，不需要知道我们正在想什么。这意味着人工智能系统并不能直接“看见”我们的大脑内或者我们正在描绘的事物，它只需要借助脑波（磁共振成像数据）来猜测我们在想什么，并从中提取图像。例如，一个字母或形状。

为了训练这项技术，研究人员在向实验者展示图像后，用记录下来的脑波给他们训练。在大约10周的时间里，他们实时收集大脑活动数据，同时把他们看到的实物进行了可视化处理。

这项“读取”我们想法的AI技术仍有较大的提升空间。我们相信，在不久的将来，一些更先进的系统，可以为新的通信方式提供解决办法。届时，我们都不需要说话甚至做手势就可以完成交流。类似的应用会有很多，例如，这项技术可以用来帮助说话有困难的人“说话”。

原文：

Certainly, the capabilities of artificial intelligence are far higher than just poweringvoice assistants and robots making backflips. Humans have always wanted sometech that could read their minds. Multiple attempts have been made so far. Andwith the start of the year 2018, here is another one.

Scientists at the ATR Computational Neuroscience Labs in Japan have created an AI-basedsystem that’s capable of performing deep image reconstruction from human brainactivity.

In simple words, it’s reading our minds without actually knowing what we’rethinking. Now, what that means is the AI system can’t see inside our brain orthe things we’re picturing. It takes help of the brain waves (MRI data) toguess what we are thinking and draws an image out of it. For instance, a letteror shape.

To train their AI, the researchers fed it with recorded brainwaves of humansubjects after showing them images. Over the course of 10 weeks, they collectedbrain activity data in real-time and also by making the human subjectsvisualize what they had seen in the past.

There is still scope for improvement for the AI which lacks perfection when it comesto reading our minds. But in the future, more advanced systems could open waysfor new modes of communication. Possibly, which won’t require us to speak oreven make hand gestures. There could be many applications. For example, suchtechnology could be used to help people with difficulty in speaking.