Google’s artificial intelligence has found how to encrypt the message of itself

Whether in the future we have turned into the destruction Robot as in the terminator thanks to the Skynet AI mischievous? Best website for latest technology news informed that Google ‘s artificial intelligence has found how to encrypt the message of itself.

You read correctly. Neural networks of artificial intelligence can think about the encryption method to secretly send messages to each other. Two researchers of Google Brain announced that they had succeeded in creating a script without any human intervention. Instead of the whole gray matter of human intelligence create, is now being used to protect the conversations between AI – and their owners – from eavesdropping.

google-ai-1

Based on the frequency with AI or chatbot decided that it will wipe out the human race or it simply not want to kill them for doing a zoo person. Then you should start to feel panic and quickly go seek shelter.

google-ai-2

But before you packed your things and go away, just calm down. Although machines now have the ability to communicate with each other by using the encrypted conversations that we do not understand. The reality that we stand in the accounting world and subjugated people as in the film Terminator’s destruction is too difficult to happen.

Google Brain’s experiment

In an experiment, two of Google’s researchers Martin Abadi and David G. Andersen have created three artificial neural networks. They respectively named Alice, Bob and Eve. Then test posed as Alice sent a letter to Bob to encode information in the mail. Then Eve tries to decipher. Turn each AI are emailing and build a secure method for content inside it. To ensure the security of the content, Alice has shifted the terms of an obscure passage to anyone. It includes which Eve could not understand. Bob can only guess the obscure content that Alice had sent. While both of them had defined with together a number of key recommendations while Eve has not permission to access.

google-ai-3

Basically, the security information in the message system of artificial neural network is rather poor. Evidence that the initial test with a simple arithmetic algorithm of Alice and Bob have been Eve catch all. However, when it is practiced repeatedly, the researchers noticed that Alice has gradually developed strategies of their own encryption. And Bob also began forming effective way of decoding. The team said they have repeated this experiment 15,000 times. And Bob may transfer all obscure passages of Alice sent the text means. Meanwhile, Eve is also learned how it can decode and 8 on 16-bit predictable from the message. But every bit as there are only 2 values 0 or 1 so this ratio is still 50%, similar to a coin toss.

google-ai-4

Result

Unexpected result especially when Alice and Bob were able to develop their own encryption methods quickly, continuously forming tactical chemical encoding and decoding that only they can understand. Even if Eve is still trying groping.

Therefore, although the content of the conversation between Alice and Bob secretly far can grow into a network Skynet. It has proved that artificial intelligence has completely succeeded in manipulating an encryption method that we developed ourselves. And the other AI also cannot be solved.

google-ai-5

Unfortunately, the team still could not find the algorithm of this method. Even by machine learning simply give one answer, not for specific know how we did. This also means that its commercial application is relatively limited, because of the current security system cannot work well with other algorithms. In short, we do not gain any benefit from the AI learned how to encrypt.

AI artificial intelligence Best website for latest technology news Google Google Brain Robot Skynet AI terminator

Related Post