Microsoft has created a bot that can generate comments under news articles. AI, known as DeepCom, was developed by a team of engineers from Microsoft and Beihan University in China.

“Automatically creating news comments is useful for real-world applications, but so far has not attracted enough attention from the research community,” the authors of the bot said in an article published on arXiv late last month.

The article did not mention the potential use of this technology to the detriment. However, as The Register notes, DeepCom has many potential flaws. For example, repressive regimes may use such a model to conduct propaganda. Fake generated comments can also lead to debate between bots and people and ultimately sow discord and misinformation. Perhaps attackers can even use the bot as a way to advertise products or to place spam. Basically, some experts say, this is a trolling machine.

DeepCom uses two neural networks: a reading network and a generation network. The reading network processes various parts of the article, starting with its title and then its contents, to analyze and predict which moments in the story are especially important or interesting. These forecasts are then transmitted to the generation network. Here, the model processes and encodes the answers, focusing on a specific topic or person of interest in the article, and decodes what it generated back into words to form comments.

habr.com