Some people may think it is too exaggerated to talk about this topic,
The first memories of what happened in the history of human beings what incredible things.
Unbelievable things, need to take some riders to determine.
We asked a was born in the year 2000 was born (Han) through to 1600 A.D. (Ming Dynasty), despite crossing the 1600 years, but the man may to around people's life will not feel too exaggerated, but for several dynasties, still live a facing loess back upturned, bale.
But if the British, who had been in the UK for 1 1600, crossed to England in 1850 to see the huge steel monster running around in the water, the man might have been scared straight away, a man 250 years ago.
If we get a person from 1850 through to 1980, I heard that a bomb can be razed to the ground by a city, the person may directly scared silly, 130 years ago Nobel is not invented dynamite.
Then 1 more people in 1980 to the present? This person will not be scared?
If 35 years ago, almost completely unable to imagine the life of the Internet age, then the human civilization into the development of the index today, how can we imagine 35 years after the era?
Super artificial intelligence, is 35 years after the ruler.
First, we define the classification of artificial intelligence:
The current mainstream view of the classification is three.
Weak artificial intelligence: weak artificial intelligence is good at a single aspect of artificial intelligence. For example, Alfa dog, can overcome the human in the game, but you have to ask him Li Shishi and Ke Jie who is more handsome, he can not answer.
Weak artificial intelligence depends on the computer's powerful computing power and repetitive logic, seemingly smart, in fact, can only do some of the precision of manual work.
At present, there are many in the automotive production line is weak artificial intelligence, so in the era of weak artificial intelligence development, human beings are indeed usher in a number of unemployed tide, but also to explore a lot of new industries.
Strong artificial intelligence: the human level of artificial intelligence. Strong AI is artificial intelligence in various aspects can emulate the human and human, capable of living brain are capable of it. Create strong artificial intelligence is much more difficult than to create a weak artificial intelligence.
Baidu Baidu's brain and Microsoft wheatgrass, are regarded as strong artificial intelligence to explore, through the huge data help strong artificial intelligence learning gradually.
The arrival of the era of strong artificial intelligence, human beings will have a lot of new fun, there will be a lot of new moral concepts.
Super artificial intelligence: all aspects of artificial intelligence more than humans. Super artificial intelligence can be a little stronger than the human, but also can be the existence of all aspects of the human strong trillion times.
When artificial intelligence learning and self correction, will continue to accelerate learning. This process may will have a sense of self, may not produce self consciousness, only to be sure is his ability will be greatly improved, which including the ability to create (alpha dog will according to the player's the fewest adjustment strategy is the most superficial layer of innovation is reflected, the ordinary mobile phone version of the game, the fewest computer actually fixed several).
We are far from the super artificial intelligence era, in the end how far it?
The first is the computing power of the computer,
Computer computing power doubling every two years, which is supported by historical data. At present, the capabilities of the human brain is 10^16 CPS, which is 1 quadrillion calculations per second. Now the fastest supercomputer, China's Milky way two, in fact, has more than this computing power.
At present, our average person to buy computer computing power is only equivalent to the level of 1/1000 of the human brain. Sounds like a weak burst, but according to the current development of electronic equipment speed, we spend 5000 yuan in 2025 to buy and the human brain computing speed to compete with the computer.
The second is to make computers more intelligent,
At present, there are two kinds of attempts to make computers become intelligent, one is to do research. Now, we have been able to simulate 1 mm long flatworms of the brain, the brain contains 302 neurons. The human brain has 100 billion neurons, and it sounds very far away. But to remember the power of exponential growth -- we have to small insect's brain, a simulation, the brains of the ants is not far away, then is the brains of rats, then simulated the human brain is not so unrealistic things.
The other is to imitate the learning process, so that the artificial intelligence constantly revised. Based on the huge data generated by the Internet, so that artificial intelligence constantly learning new things, and constantly self correction. Baidu's Baidu brain is said to have 4 years of age, can be a continuous dialogue, you can judge a person's action according to the picture. Although the number is still a lot of mistakes, but this change ability is a qualitative change.
In the eyes of the world's brightest scientists, the emergence of strong artificial intelligence is no longer a problem will not, but the problem of when, in 2013, a hundreds of artificial intelligence experts to participate in the survey do you predict human level artificial intelligence, when will come true? "
The results are as follows:
2030: 42% of respondents believe that strong artificial intelligence will be achieved
2050: 25% of respondents
2070: 20%
After 2070: 10%
Never realized: 2%
That is, more than 2/3 of scientists think that 2050 years ago, the strong artificial intelligence will be realized, while only 2% of people think that it will never be achieved
his situation has never happened in history.
Kurzweil is president of the singularity University (Google, the US National Aeronautics and Space Administration and a number of technical experts to jointly establish) with more optimistic estimates, he believed that computers would be a strong artificial intelligence in 2029, by 2045, enter the era of super artificial intelligence.
So, if you think you can live for 30 or 40 years, then you should be able to witness the emergence of super artificial intelligence.翻譯：
人工智能來臨，有人在擔憂失業，有人在憧憬未來，有人在發掘行業機會，也有人在研究圍棋。在討論這些之前，也許我們應該先考慮壹下人類的結局。
有人可能覺得談論這個話題太誇張了，
那先回憶壹下人類歷史上究竟發生了哪些不可思議的事情。
不可思議的事情，需要請幾個穿越者來判定。
我們請1個出生於公元0年出生的人（漢朝人）穿越到公元1600年（明朝），盡管跨越了1600年，但這個人可能對周圍人的生活不會感到太誇張，只不過換了幾個王朝，依舊過著面朝黃土背朝天的日子罷了。
但如果請1個1600年的英國人穿越到1850年的英國，看到巨大的鋼鐵怪物在水上路上跑來跑去，這個人可能直接被嚇尿了，這是250年前的人從未想象過的。
如果再請1個1850的人穿越到1980年，聽說壹顆炸彈可以夷平壹座城市，這個人可能直接嚇傻了，130年前諾貝爾都還沒有發明出炸藥。
那再請1個1980年的人到現在呢？這個人會不會被嚇哭呢？
如果35年前的人，幾乎完全無法想象互聯網時代的生活，那麽人類文明進入指數發展的今天，我們怎麽能想象35年後的時代？
超人工智能，則是35年後的統治者。
首先，我們明確壹下人工智能的分類：
目前主流觀點的分類是三種。
弱人工智能：弱人工智能是擅長於單個方面的人工智能。比如阿爾法狗，能夠在圍棋方面戰勝人類，但妳要問他李世石和柯潔誰更帥，他就無法回答了。
弱人工智能依賴於計算機強大的運算能力和重復性的邏輯，看似聰明，其實只能做壹些精密的體力活。
目前在汽車生產線上就有很多是弱人工智能，所以在弱人工智能發展的時代，人類確實會迎來壹批失業潮，也會發掘出很多新行業。
強人工智能：人類級別的人工智能。強人工智能是指在各方面都能和人類比肩的人工智能，人類能幹的腦力活它都能幹。創造強人工智能比創造弱人工智能難得多。
百度的百度大腦和微軟的小冰，都算是往強人工智能的探索，通過龐大的數據，幫助強人工智能逐漸學習。
強人工智能時代的到來，人類會有很多新的樂趣，也會有很多新的道德觀念。
超人工智能：各方面都超過人類的人工智能。超人工智能可以是各方面都比人類強壹點，也可以是各方面都比人類強萬億倍的存在。
當人工智能學會學習和自我糾錯之後，會不斷加速學習，這個過程可能會產生自我意識，可能不會產生自我意識，唯壹可以肯定的是他的能力會得到極大的提高，這其中包括創造能力（阿爾法狗會根據棋手的棋路調整策略就是最淺層的創新體現，普通手機版的圍棋，電腦棋路其實就固定的幾種）。
我們距離超人工智能時代，到底有多遠呢？
首先是電腦的運算能力，
電腦運算能力每兩年就翻壹倍，這是有歷史數據支撐的。目前人腦的運算能力是10^16 cps，也就是1億億次計算每秒。現在最快的超級計算機，中國的天河二號，其實已經超過這個運算力了。
而目前我們普通人買的電腦運算能力只相當於人腦千分之壹的水平。聽起來還是弱爆了，但是，按照目前電子設備的發展速度，我們在2025年花5000人民幣就可以買到和人腦運算速度抗衡的電腦了。
其次是讓電腦變得智能，
目前有兩種嘗試讓電腦變得智能，壹種是做類腦研究。現在，我們已經能夠模擬1毫米長的扁蟲的大腦，這個大腦含有302個神經元。人類的大腦有1000億個神經元，聽起來還差很遠。但是要記住指數增長的威力——我們已經能模擬小蟲子的大腦了，螞蟻的大腦也不遠了，接著就是老鼠的大腦，到那時模擬人類大腦就不是那麽不現實的事情了。
另壹種是模仿學習過程，讓人工智能不斷修正。基於互聯網產生的龐大數據，讓人工智能不斷學習新的東西，並且不斷進行自我更正。百度的百度大腦據說目前有4歲的智力，可以進行幾段連續的對話，可以根據圖片判斷壹個人的動作。盡管目前出錯的次數依舊很多，但是這種能力的變化是壹種質變。
在全球最聰明的科學家眼中，強人工智能的出現已經不再是會不會的問題，而是什麽時候的問題，2013年，有壹個數百位人工智能專家參與的調查 “妳預測人類級別的強人工智能什麽時候會實現？”
結果如下：
2030年：42%的回答者認為強人工智能會實現
2050年：25%的回答者
2070年：20%
2070年以後：10%
永遠不會實現：2%
也就是說，超過2/3的科學家的科學家認為2050年前強人工智能就會實現，而只有2%的人認為它永遠不會實現。
最關鍵的是，全球最頂尖的精英正在拋棄互聯網，轉向人工智能——斯坦福、麻省理工、卡內基梅隆、伯克利四所名校人工智能專業的博士生第壹份offer已經可以拿到200-300萬美金。這種情況歷史上從來沒有發生過。
奇點大學（谷歌、美國國家航天航空局以及若幹科技界專家聯合建立）的校長庫茲韋爾則抱有更樂觀的估計，他相信電腦會在2029年達成強人工智能，到2045年，進入超人工智能時代。
所以，如果妳覺得妳還能活30、40年的話，那妳應該能見證超人工智能的出現。