While artificial intelligence is an incredibly important field that’s growing by leaps and bounds, perhaps its most interesting lessons concerns just how incredible the human brain is at performing certain functions. While computers might be better at performing math and looking dozens of chess moves into the future, they can’t yet compete with the human brain at figuring out things like a video’s topic.
A recent research project demonstrated just that fact by feeding videos to Google’s Cloud Video Intelligence API and seeing if it could determine exactly what a given video was about. Apparently, this seemingly simple task is a challenge for Google’s AI and points out the difficulty of creating automatic systems to categorize video, as Motherboard reports.
The research team in question works at the University of Washington, and the team used some trickery to see how smart the Google API really is. Currently in beta, the Google Cloud Video Intelligence API has one job, which was to “make video searchable” and to annotate video to make it easier for humans to search through them.
In their tests, the researchers injected extraneous, and subliminal, images of a pasta bowl into a video featuring primatologist Jane Goodall and gorillas. The result was that the Google AI concluded that the video was actually about spaghetti and not the apes. Another example involved placing a picture of an Audi into a video about tigers, which caused the AI to conclude that the video was about cars.
Although it might sound somewhat comical, these mistakes point out a serious issue with the AI. As the researchers noted in their conclusion:
“However, we showed that the API has certain security weaknesses. Specifically, an adversary can insert an image, periodically and at a very low rate, into the video in a way that all the generated shot labels are about the inserted image. Such vulnerability seriously undermines the applicability of the API in adversarial environments.”
Even worse, according to the researchers, “Furthermore, an adversary can bypass a video filtering system by inserting a benign image into a video with illegal contents.” The fact that the process of doing so requires no specialized knowledge about the AI’s machine learning algorithms or about video annotation in general was particularly disturbing.
Ultimately, what the research points out is that AI has a long way to go before it can match the human brain in determining things like a video’s topic. Inserting subliminal messages into video has been known for a long time to affect the human psyche, but at least a human wouldn’t think that a video about apes is actually about spaghetti — the human would probably just start craving pasta instead.
