After Microsoft Tay Hacked, Facebook robots should do a good job of protecting

Now the robot is not going to fire. It seems that the entire high-tech industry is developing an AI-based assistant in the application to do simple, trivial tasks for you, saving you time and making your work more efficient. Just this month, Microsoft and Facebook have released developer tools that make it easier to build robots on their platforms.


Given the current size of the robot, like Microsoft and Facebook and the company wants to see it, it's reasonable to worry about how much of this technology might be supervised there.


This wave of interest in robotics does usher in a new economy, internal messaging platforms and search engine stealth assistants vying for our feelings. For now, its use may seem commonplace-ordering pizza by text message or buying a burrito in a mobile app. But as chat bots increase and become more proficient, they can provide a smarter, more personalized service experience. Imagine a robot answering your questions about legal matters, even your mental health.


Therefore, programmers must remain cautious. Robots that are not properly trained and monitored can become ugly. Don't worry, there is no danger of any robot uprising, but there have been robots that deceive humans, and it is well known that Microsoft's Tay is blackened. High-tech companies must keep these things in mind when developing dialogue robots.


Spontaneous behavior


As an artificial intelligent robot, in addition to programmed, it can learn new skills and perform other tasks. Robot experts call it spontaneous behavior. Robots often do not have human filters or guides when they are aware of new tasks. This is the chat robot Tay, originally programmed to learn from the dialogue, and finally turned into a hatred of the Nazi reason. To prevent this from happening again, the key is to proactively monitor it.


"I'll check all my bots, but more, I'll think if the input stuff will produce problematic content. Norarid says it has created many popular Twitter robots. "I'm going to avoid something, and I don't want these things to be put together with others because it leads to marginalization or paranoia. ”


Given the scale of the current deployment of robots, such as Microsoft, Facebook and robotics, it is reasonable to worry about whether enough people are monitoring the technology. Chenglili, manager of the future social testing lab at Microsoft, said the company is considering "other types of tools that we can help to get better hints when developers have problems with their robot behavior." "The developers will want to know if their bots are sending inappropriate content, or if there are other anomalies," the process says.


Even after the post-mortem analysis, the robot programmer should improve the robot so that they no longer have trouble with spontaneous behavior. To do this, you can restrict the subject to get data, or make a blacklist of inappropriate words for robots. These are common practices that Chat robot interaction designer Caroline de says

After Microsoft Tay Hacked, Facebook robots should do a good job of protecting