Whether artificial intelligence is suitable for the field of security.
Following Google's announcement of the breakthrough research of its own intelligence project AlphaGo - the first time a computer program defeated an amateur Go player, Google DeepMind announced the detailed environment for the final challenge between Lee Sedol, the best Go player of the past decade. From March 9th to March 15th, AlphaGo will hold 5 challenge matches against Lee Sedol in Seoul, South Korea. The competition is completely equal, and the winner will receive a $1 million prize.
In today's battle, the first match between Google's domestic intelligence AlphaGo and South Korean chess player Lee Sedol ended, and AlphaGo won today's match. After 3 and a half hours of competition, Lee Sedol conceded defeat. After the game stops today, the two sides will also leave for the remaining 4 games on the 10th, 12th, 13th, and 15th at 12 midnight Beijing time.
DeepMind, the pioneer of AlphaGo, published a cover article in the academic journal Nature in January this year, showing the details of the breakthrough in the field of Go home intelligence. In this chess battle, sitting on the face of Li Shishi was Aja Huang, one of the authors of the cover article of "Nature", who himself replaced AlphaGo on the chessboard. The flag of the AlphaGo side is the British flag, because the google Go sequence AlphaGo in this competition comes from the British domestic intelligence company DeepMind.
Google's AlphaGo won its first match against the world's top Go player, a milestone in the history of domestic intelligence development, indicating that domestic intelligence has been able to perform beyond humans in highly complex games such as Go.
How powerful is Google's AlphaGo
The traditional domestic intelligence method is to build a search tree from the entire approximate move, but this kind of method is not suitable for Go. The AlphaGo launched by Google this time combines advanced search trees and deep neural networks. These neural networks transmit descriptions of the board through 12 fix layers, which contain millions of neural-like connections.
One neural network "policy network" is responsible for choosing the next move, and the other neural network "value network" predicts the winner of the game. Google uses 30,000 moves of a human Go player. Go moves to train the neural network, and at the same time, AlphaGo is also researching new strategies on its own, running thousands of games of Go between its neural networks, manipulating frequent trials to mediate connection points, a process also known as reinforcement learning. , through the popular use of Google Cloud Platform, to achieve a small number of research tasks.
AlphaGo is fighting a human chess game with "two brains" to get the title: "Decision Collection" and "Value Collection". In layman's terms, one brain is used to decide what to do in the future, and the other is used to predict the ultimate winner of the competition.
Google hopes to use this family of intelligent algorithms for disaster prediction, injury control, medical health, and death boards and other complex fields. That is to say, the purpose of google is still to serve human beings better, without any malice. Hassabis, founder of DeepMind (AlphaGo's research company), said: "Private warnings about domestic intelligence overshadow the help that domestic intelligence brings. Domestic intelligence at the human brain level is still quite far away, and it will take about ten years.
In yesterday's pre-match announcement meeting, Google Chairman Schmidt said that winning or losing is a victory for mankind. Because it is the happiness of human beings that makes the breakthrough of domestic intelligence at this moment.
In August this year, a network challenge competition called Darpa's Cyber Grand Challenge will be held. The seven teams participating in the final must attack the opponent's gap while seeing and automatically repairing their own gaps, shielding functions and individual functions. Unlike other "capture the flag" competitions, this is the world's first hacking competition entirely dominated by computer software.
The winner will initially win the $2 million grand prize, Darpa is actually the US Defense Advanced Research Projects Agency, no wonder people have deep pockets!
Giovanni Vigna, a professor of computer science at the University of California, Santa Barbara, is the founder of the hacker team "Shellphish", and they set up a system called "Mechanical Phish" (machine crawl) to participate in this competition.
Vigna said at the 2016 RSA Peace Conference in the US:
"A fully automated hacking system is our ultimate frontier. Humans can see the gaps, but can't analyze millions of projects."
Hackers in practice are actually not as "sexy" as the film depicts, he said:
"A hacker is just a bunch of people around a table, tired but still typing on the keyboard on a notebook. And we are hackers either because we want to hit people, or we want to find Lu The flaws are probably just because it is easy to play."
"Death board hackers" perform well in the process of protecting the network, rapid identification, system patching, etc., thereby avoiding the control of loopholes, theft of data or the attack of network services.
Household Intelligence Quickly Clears "Super Mary"
In addition to the Cyber Grand Challenge, other organizations have also begun to apply domestic intelligence to hacking death boards.
Konstantinos Karagiannis built a hacking system with neural networks (simulating the human brain to learn and get things done). According to him, this domestic smart program called MarI/O can pass "Super Mary" in only 34 tests without any prior knowledge. In the end, it only has a number of complex parameter configurations, and after a number of different tests, it "understands" how to play, and it "learns".
"A calm scanner with AI technology can identify cluttered crevices in a way that people never imagined. It can use very low-end hardware, a thousand-dollar GPU (graphics processing unit, Often in games) can outperform a supercomputer from a decade ago."
Karagiannis hopes to demo this POC this summer.
Every coin has another side
HackerOne security co-founder Alex Rice agrees with the concept, arguing that any preventive technology that can be used to see the gaps is likely to be in the hands of criminals - the technological ultimate city becomes a double-edged sword.
Nonetheless, Rice sees the rise of automated calming methods as a virtue:
"Everyone is happily following the trend. Almost the entire structure has experienced cyber attacks, it's about life and death, obviously everything we've done before has been successful. And the best way to do this is to connect humans with death. We have yet to see the tools that humans are best at. Even if we have sentient death boards in the future, they will still be instructed by humans.”