China could fight porn... by reading minds

China could fight porn… by reading minds

To lend a hand to the censors, China may be able to count on neuroengineering.

China is one of the countries waging an all-out war on open-access pornography. On the Chinese Internet, even the slightest explicit image is actively censored by a special division of the government. And the latter could one day benefit from a reinforcement that is surprising to say the least; Chinese researchers have announced that they have developed a technology capable of “read the mind” of a subject to determine if they are exposed to pornographic images.

The concept might seem straight out of a sadly generic sci-fi movie, but it’s actually very real. It is mostly based on artificial intelligence, a technology that works wonders when it comes to image processing.

Starting from existing technologies which are already very advanced, one can therefore quite imagine developing a system capable of automatically identifying all images of a pornographic nature.

The problem is that human criteria for decency are sometimes quite vague and vary widely across cultures. Moreover, it is common knowledge that these computer systems also have their limits; when these algorithms are confronted with situations that they do not know, they can provide completely fanciful answers.

A pornography detector helmet at the service of censors

For these reasons, it is still complicated to leave the keys of the truck to the computer, especially when the responsibility of a government institution is engaged. Based on this observation, Chinese researchers have therefore developed a system to put the power of AI at the service of a human in the flesh.

It takes the form of a kind of flexible helmet covered with electrodes. These latter allow capture brain waves of the bearer, in this case a censor. These are interesting elements, because they are visual representations of neurological phenomena that take place completely unconsciously.

The researchers therefore began by training their artificial intelligence by having several subjects wear this helmet. They were subjected to a stream of varied images, some of which were pornographic in nature and generated particular brain wave patterns.

From these readings, they were able to train the AI ​​to specifically spot these patterns; so she is able to determine in real time whether the user is viewing sexually explicit images. And this even if he only lays his eyes on it for a fraction of a second. The rest of the readings were used to build a second model, which this time makes it possible to filter out human errors, in particular those related to fatigue and emotions.

A concept still limited but with immense potential

On paper, this system therefore makes it possible to take advantage of the respective qualities of the human and the machine to achieve an extremely fast and powerful system. At the end of the training process, the system performed quite well, with more than 80% accuracy during identification. But the researchers also explain that their work has several important limitations.

To begin with, they claim to have had great difficulty in find source material – implied, uncensored pornographic images. No wonder since they had to limit themselves to the Chinese Internet which is extremely controlled at this level. But it is nevertheless a big problem in this kind of work.

Indeed, the quality of the results provided by an AI-based system directly depends on the size and quality of the dataset. If they are insufficient, we find ourselves in a scenario called “garbage in, garbage out” (“waste in, waste out”) where the results are no longer worth much.

Additionally, the researchers were only able to test their system on male subjects. This is an already significant limit in normal times which is all the more restrictive in this type of study in connection with pornography, especially since a large proportion of government censors are women, according to the South China Morning Post.

neuralink-elon-musk
Like Elon Musk’s Neuralink, all of these new neuroengineering systems will need to be handled with care. © Neuralink

Important issues in the age of AI and neurobusiness

Alongside the societal questions about access to pornography that we will not address here, there is also a whole train of ethical questions. Because according to a researcher interviewed by Interesting Engineering on condition of anonymity, there is for the moment no tools or laws to regulate how these brain waves will be used.

And this is still a big concern. Because for the moment, pno one yet knows the extent of the potential of this technology emergent whose uses will soon go far beyond censorship. One can imagine a simplistic example to illustrate the problem. Knowing the complicated relationship between the Chinese government and the LGBT community, one can for example imagine that a censor could be put in a very uncomfortable situation if his brain waves betray a particular reaction to certain types of content.

This is just an isolated and perhaps far-fetched example, but it illustrates one of the main problems associated with systems that allow “mind reading”. Suffice to say that if the technological bases are interesting, it will be necessary remain particularly vigilant at a time when neuro-engineering and artificial intelligence are gaining momentum at breakneck speed.

Leave a Comment

Your email address will not be published.