Thursday, August 31, 2017

Google’s Deep Mind and The Future of Artificial Intelligence

Deep mind was a startup founded in 2010. Their goal is to “solve intelligence” through machine learning and mimicking how the human mind works by building a self-learning algorithm. The deep mind project is attempting to create a computer program that is completely self-taught and eventually will have the ability to solve complex problems without human intervention.
Artificial Intelligence or AI has been a topic of discussion ever since the invention of computers and in the past decade or so the creation of AI has become more and more of a reality. Googles deep mind has had a recent success by successfully being creating a program called Alpha Go. The purpose of this program is to have the program successfully beat a grandmaster Go player in the worlds hardest and most complex game called Go.
4-1-google-deepmind-beats-go-champion-lee-sedol-in-a-tense-final-game
Go is an ancient Chinese board game that has been played for centuries and is the most complicated board game in the world. This is not the first time a computer has been able to beat a grandmaster in a board game. We have seen this done before in the game chess in the chess example what was done was to input all possible outcomes from of game of chess and from that the computer could beat even the best player by always selecting the most optimal solution. What makes the fact that the program was able to beat a Go grandmaster so significant is the sheer complexity of the game itself. The game of has as many possible moves to chess as there are atoms in the entire universe. Where it was possible in chess to map all possible outcomes in the game the complexity of the game Go makes mapping out all outcomes essentially impossible.
So if it’s impossible to code in every possible outcome into to the program how did a computer beat the Grand Master? The Alpha Go program was taught how to play the game of Go by exposing it to hundreds of thousands of games. By exposing it to all of these games gave it the ability to learn how the game worked. After the program learned from all of these games it had the same skill set as an amateur player by being able to mimic how all of those players were able to play. But the goal is not to mimic the players its goal is to become the best player in the game so in order to do this they had the program play itself in the game over 30 million times. By playing it learns from its own mistakes and is able to incrementally grow better every game. By learning from its own mistakes and playing itself it makes the program learn how to play better. Compounding this self-learning process over 30 million times gives it the ability to beat even the best of players.
What Makes This So Significant?
This program isn’t designed just to play the game of Go, it is possible to apply this same self-learning technique to any problem that you expose it to. Deep Mind is now being used in industries such as scientific research, mathematics, finance, and healthcare and through this machine learning technique it can solve any number of problems that Humans haven’t been able to solve for themselves. This will lead to an infinite amount of scientific breakthroughs and Deep Dream will be able to solve problems that mankind would have never been able to solve for itself. The possibilities are endless and self-learning AI is proving to be one of the biggest technological breakthroughs in mankind. Through machine learning, we will be soon be experiencing exponential growth in technology beyond our imagination.

6 comments:

  1. This comment has been removed by the author.

    ReplyDelete
  2. Interesting post. I recently read an article about an AI program beating two professional eSports players only after two weeks of practice. According to the article, the game, Dota 2, is known to be extremely complex, and engineers of the program say that it got “lifetimes of experience” in just two weeks of learning. I think it’s remarkable that AI can learn so quickly to beat people in games like chess, Go, Dota 2, etc. I wonder if the time will come when AI can surpass humans and threaten human existence as Elon Musk and Stephen Hawking warn.

    https://www.theverge.com/2017/8/11/16137388/dota-2-dendi-open-ai-elon-musk

    ReplyDelete
    Replies
    1. Dota 2 professional eSports player being beat by Elon Musk's Open AI is incredible in terms of the progress of strategic AI. The fact that AI has gone from beating humans in chess to Dota 2 shows the ability of the AI to evolve and adapt to fit more of a human like play style. I appreciate the link Takeshi.

      There is a software firm founded in America in 2012 that was been named the “first company to apply artificial intelligence, algorithmic science and machine learning to cyber security”.1 Cylance Inc. is the name of this company. They claim a predictive security measure, vs a retroactive approach. AI has come so far, and it has spread to many fields of business. Just goes to show that with this software, human resources can be directed to more demanding tasks-as predictive means less recovery efforts if a security breach occurs.

      1: Guru, Sagganik; Khyati, Jain (May 12, 2016). "U.S. based Cylance Inc. and Japan’s MOTEX Unite to Establish Advanced Cyber Threat Prevention Solutions in Japan". Cyber Secure India.

      Delete
  3. The article seems extremely relevant today, as I just saw that Vladmir Putin expressed that he believed that the country with the best AI in the coming years is going to be the country that rules the world. Elon Musk has also been trying to preach about the dangers of AI lately. He believes that it is more dangerous than the North Korean threat.

    ReplyDelete
  4. What I am interested in seeing is the effect that artificial intelligence has on the job market in the future. Wether or not a universal income has to be involved is an interesting topic that industry leaders such as Musk have addressed.

    ReplyDelete
    Replies
    1. Based on a talk by Mike Munger1, artificial intelligence may be referred to as a software killer. As in, software, like AI, kills service jobs by way of taking the jobs of those in positions like cashier or ticket seller. If a software came out that could do the job of the person who gives you your ticket at an airport, then this persons job is at stake because the software may be cheaper to employ. Thus, software killer.

      Mike Munger also goes into the interesting topic of Universal income1. Perhaps universal income will be employed so those at risk of employment loss due to automation of both industry and service jobs can reach out to other job sectors; maybe by way of education systems to get them into the newer job sectors.

      1: http://www.econtalk.org/archives/2017/01/michael_munger_3.html

      Delete