Google the leading search engine platform is yet to make another achievement. The researchers from Google Brain Deep Project who is already into making its AI learn works of art are now set to enable the AI to encrypt messages independently without any human intervention.
As per the new research paper published Googlers Martín Abadi and David G. Andersen have voluntarily allowed the three AI test subjects — neural networks named Alice, Bob and Eve to pass each other’s notes which are encrypted by AIs themselves.
According to the reports published in New Scientist, Abadi and Andersen allotted each AI a task: Alice had to send a secret message that only Bob could read, while Eve shall try to figure out how to eavesdrop and decode the message herself.
The experiment began with a plain text message which Alice converted it into an unreadable language. Bob was able to decode the using cipher key. Initially, Alice and Bob were bad at hiding their secrets. Nevertheless, gradually over the course of 15,000 attempts, Alice was able to come up with her own encryption strategy. While Bob has figured out to decrypt such messages.
The message consisted of just 16 bits and was in binary language, which means it included either 1 or 0. Eve on the other end able to guess half of the bits in the message which denotes that she was she was basically just flipping a coin or guessing at random.
The researchers do not yet know the kind of encryption method used by Alice. So there are no practical applications to decipher from it. The scientists claim that personifying the AI oversimplifies things because of the way machine learning works.
With open-source deep learning tools like Microsoft’s Cognitive Toolkit, it might be interesting to see this work on an even larger scale.