[Skynet] Google's neural network is learning to encrypt its own messages

[Hamartia Antidote]

http://www.dailymail.co.uk/sciencet...neural-network-learning-encrypt-messages.html

Experts like Professor Stephen Hawking and Elon Musk have warned of the dangers of artificial intelligence becoming too smart and turning against humanity.

Now it seems a team at Google has brought computing another step towards this nightmare becoming a reality, by teaching its networks to keep secrets.

The computer systems have learn how to protect their messages away from prying eyes.

Google Brain team, Martin Abadi and David Andersen, published a paper showing how computer networks can work out how to use a simple encryption technique.

The Google research group specialises in the area of deep learning, using huge amounts of data.

The computing system was designed based on the way neurons work in the brain, a type of computer called a 'neural network'.

In a paper published on the preprint server arXiv, the researchers said:

'We ask whether neural networks can learn to use secret keys to protect information from other neural networks.

'Specifically, we focus on ensuring confidentiality properties in a multiagent system, and we specify those properties in terms of an adversary.'

The team created a system that 'may consist of neural networks named Alice and Bob, and we aim to limit what a third neural network named Eve learns from eavesdropping on the communication between Alice and Bob,' they said.

Alice, Bob and Eve are names computer scientists and physicists always use to name parts of their experiments.

To make sure the message remained secret, the network named Alice had to convert her original plain-text message into something unintelligible.

This meant anyone who intercepted it, including Eve, would not be able to understand the message.

This 'cipher text' had to be decipherable by Bob, but nobody else.

At first the networks were not good at hiding the messages, but after some practice they got better and were able to hide the messages from Eve.

This is a big step, the researchers said, because neural networks, or 'neural nets', were not designed to be good at encryption.

'Computing with neural nets on this scale has only become possible in the last few years, so we really are at the beginning of what's possible,' Joe Sturonas of encryption company PKWARE in Milwaukee, who was not involved in the research, told New Scientist.

Just last week, Professor Stephen Hawking warned artificial intelligence could develop a will of its own that is in conflict with that of humanity.

It could herald dangers like powerful autonomous weapons and ways for the few to oppress the many, he said, as he called for more research in the area.

But if sufficient research is done to avoid the risks, it could help in humanity's aims to 'finally eradicate disease and poverty', he added.

 

[Echo_419]

http://www.dailymail.co.uk/sciencet...neural-network-learning-encrypt-messages.html

Experts like Professor Stephen Hawking and Elon Musk have warned of the dangers of artificial intelligence becoming too smart and turning against humanity.

Now it seems a team at Google has brought computing another step towards this nightmare becoming a reality, by teaching its networks to keep secrets.

The computer systems have learn how to protect their messages away from prying eyes.

Google Brain team, Martin Abadi and David Andersen, published a paper showing how computer networks can work out how to use a simple encryption technique.

The Google research group specialises in the area of deep learning, using huge amounts of data.

The computing system was designed based on the way neurons work in the brain, a type of computer called a 'neural network'.

In a paper published on the preprint server arXiv, the researchers said:

'We ask whether neural networks can learn to use secret keys to protect information from other neural networks.

'Specifically, we focus on ensuring confidentiality properties in a multiagent system, and we specify those properties in terms of an adversary.'

The team created a system that 'may consist of neural networks named Alice and Bob, and we aim to limit what a third neural network named Eve learns from eavesdropping on the communication between Alice and Bob,' they said.

Alice, Bob and Eve are names computer scientists and physicists always use to name parts of their experiments.

To make sure the message remained secret, the network named Alice had to convert her original plain-text message into something unintelligible.

This meant anyone who intercepted it, including Eve, would not be able to understand the message.

This 'cipher text' had to be decipherable by Bob, but nobody else.

At first the networks were not good at hiding the messages, but after some practice they got better and were able to hide the messages from Eve.

This is a big step, the researchers said, because neural networks, or 'neural nets', were not designed to be good at encryption.

'Computing with neural nets on this scale has only become possible in the last few years, so we really are at the beginning of what's possible,' Joe Sturonas of encryption company PKWARE in Milwaukee, who was not involved in the research, told New Scientist.

Just last week, Professor Stephen Hawking warned artificial intelligence could develop a will of its own that is in conflict with that of humanity.

It could herald dangers like powerful autonomous weapons and ways for the few to oppress the many, he said, as he called for more research in the area.

But if sufficient research is done to avoid the risks, it could help in humanity's aims to 'finally eradicate disease and poverty', he added.

These idiots will be responsible for our doom