总结过去展望未来
In 1965, a young engineer named Gordon Moore predicted that the number of transistors on an integrated circuit would double every two years.
1965年，一位名叫戈登·摩尔 ( Gordon Moore)的年轻工程师预测集成电路中晶体管的数量每两年将翻一番。
In layperson’s terms, this means that the power of computers would double every two years.
用外行的话来说，这意味着计算机的能力每两年将翻一番。
Moore made the prediction in a paper speculating about what the next ten years would hold for the nascent field of computers. In 1975, his prediction turned out to be shockingly accurate and a new law was born: Moore’s law. However, the law was even more profound than Moore’s original prediction.
摩尔在一篇论文中做出了这一预测，推测了未来十年对于新兴的计算机领域的影响。 1975年，他的预测非常精确，诞生了一条新的定律： 摩尔定律 。 但是，该法律甚至比摩尔最初的预测更为深刻。
In addition to computing power doubling every two years, the cost of that computing power would be cut in half every two years.
除了计算能力每两年翻一番，该计算能力的成本还将每两年减少一半。
And it turns out that Moore’s law has held true ever since then. Computers have doubled their computing power every two years, while the cost of all this incredible new computing power has continued to tumble. That is what has enabled all the technological marvels we’ve seen over the past few decades. Computers have gone from clunky arithmetic machines, to spreadsheets, to graphical interfaces, to video editing, to smartphones, to sequencing the human genome because of Moore’s law. All those things have required more and more computing power, and Moore’s law has delivered.
事实证明，从那时起，摩尔定律就一直成立。 计算机每两年将其计算能力提高一倍，而所有这些令人难以置信的新计算能力的成本却持续下降。 这就是过去几十年来我们看到的所有技术奇迹的原因。 由于摩尔定律，计算机已经从笨拙的运算器，电子表格，图形界面，视频编辑，智能手机到对人类基因组进行测序。 所有这些事情都需要越来越多的计算能力，摩尔定律已经付诸实践。
Moore’s law is an example of exponential growth, which is growth that occurs at a faster and faster rate over time. It is very difficult for us humans to wrap our heads around exponential growth. We’re much more accustomed to think in terms of linear growth, which is growth at a steady rate over time. When we look into the future, we tend to think in terms of linear growth.
摩尔定律就是指数增长的一个例子，即随着时间的推移，增长速度越来越快。 对于我们人类来说，要想围绕指数增长感到头疼是非常困难的。 我们习惯于线性增长，即随着时间的推移保持稳定的增长速度。 展望未来时，我们倾向于考虑线性增长。
Let’s see an example of this in practice. If I asked you to predict how much technology will change over the next ten years, you might reflect on how technology changed over the last ten years and then project that forward from the present. But that’s linear thinking. That form of thinking says the pace of development will be the same over the next ten years as it was over the last ten years.
让我们在实践中看一个例子。 如果我要求您预测未来十年技术将发生多少变化，您可能会反思过去十年中的技术变化，然后从现在开始进行预测。 但这是线性思维。 这种思维方式表示，未来十年的发展速度将与过去十年相同。
Moore’s law and exponential growth tell us that the pace of development will be faster over the next ten years than it was over the last ten years. And the pace of development over the next twenty years will be a lot faster than over the last twenty years.
摩尔定律和指数增长告诉我们，未来十年的发展速度将比过去十年更快。 未来二十年的发展速度将比过去二十年快得多。
The cutting edge of computing today is artificial intelligence.
当今计算的最前沿是人工智能。
In layperson’s terms, artificial intelligence is the field of “teaching” computers to “think” and “learn.”
用非专业人员的话来说，人工智能是“教”计算机进行“思考”和“学习”的领域。
It’s been around for a while and has probably crept into your life in some small ways. Netflix “learns” what kind of movies you like. When you use your smartphone camera, it puts a yellow box around faces and focuses on them because it “knows” they’re faces. Siri or Alexa “understand” questions you ask them.
它已经存在了一段时间，并且可能以一些小方式潜入了您的生活。 Netflix“学习”您喜欢什么样的电影。 当您使用智能手机相机时，它会在脸部周围放置一个黄色框并对其进行对焦，因为它“知道”他们是脸部。 Siri或Alexa“理解”您问的问题。
But Moore’s law has been present here as well, and the computing power has grown at a faster and faster rate. As a result, artificial intelligence software is starting to do more and more impressive and interesting things. Self-driving cars are already out on the roads, for example.
但是这里也存在摩尔定律，并且计算能力以越来越快的速度增长。 结果，人工智能软件开始做越来越多令人印象深刻和有趣的事情。 例如，无人驾驶汽车已经在路上 。
A recent example that has gained a lot of attention is a piece of software called GPT-3 that is surprisingly good at understanding and using language. Nearly 50% of people comparing one news article written by the software and one written by an actual human can’t tell who wrote which one.
一个最近引起广泛关注的例子是一个名为GPT-3的软件，它出奇的擅长理解和使用语言。 将近50％的人在比较由该软件撰写的一篇新闻文章和由实际人类撰写的一篇新闻文章时，不知道是谁撰写了哪一篇 。
And these are still the early days of artificial intelligence. Moore’s law tells us that the improvements over the next decade will be far faster and far more significant than those over the past decade.
这些仍然是人工智能的早期。 摩尔定律告诉我们，下一个十年的改进将比过去十年更快，更重要。
With artificial intelligence being the cutting edge of computing, there are naturally specialized computer chips specifically being designed for this purpose. This raises the question of whether Moore’s law is still holding true: is computing power still doubling every two years or has the pace slowed or accelerated?
随着人工智能成为计算的最前沿，自然而然地专门为此目的设计了专用的计算机芯片。 这就提出了一个问题，即摩尔定律是否仍然成立：计算能力是否仍每两年翻一番？还是步伐放慢或加快？
There’s a growing body of evidence that the pace of Moore’s law has accelerated. By a lot.
越来越多的证据表明，摩尔定律的步伐已经加快。 很多。
By some calculations, computing power for artificial intelligence is now doubling every 3.4 months. It’s hard to express how much faster than Moore’s law that is. The difference between two years and 3.4 months may not sound like a lot, but that’s linear thinking. This is exponential growth. We’re talking 50–60 times faster than Moore’s law.
通过一些计算，人工智能的计算能力现在每3.4个月翻一番 。 很难表达比摩尔定律快多少。 两年与3.4个月之间的差额听起来可能并不多，但这是线性思维。 这是指数增长。 我们谈论的速度比摩尔定律快50–60倍 。
To put some numbers on it, computing power for artificial intelligence increased 300,000 fold between 2012 and 2018. Moore’s law would have yielded an 8 fold increase.
确切地说 ，2012年至2018年间，人工智能的计算能力增加了300,000倍 。摩尔定律将产生8倍的增长。
Here’s a more concrete example. A common (and very useful) task for artificial intelligence is to classify what’s in an image. A classic test is determining whether a photo was of a dog or of a cat. A more useful implementation would be for a self-driving car to determine what’s a human, what’s a cyclist, what’s a tree, and what’s another car.
这是一个更具体的例子。 人工智能的一项常见(且非常有用)任务是对图像中的内容进行分类。 经典测试是确定照片是狗还是猫。 对于无人驾驶汽车，更有用的实现方式是确定什么是人，什么是骑自行车的人，什么是树，什么是另一辆车。
In the beginning of 2017, it cost $10,000 to classify one billion images on the latest computers. By the end of 2019 it cost 3 cents to perform the same task. That is exponential growth at a rate dramatically faster than Moore’s law.
2017年初，在最新计算机上对10亿张图像进行分类的成本为10,000美元。 到2019年底，执行同一任务的成本为3美分 。 那就是指数式增长，其速度大大快于摩尔定律。
GPT-3, the artificial intelligence language software mentioned above, is 117 times as powerful as its predecessor GPT-2. GPT-3 came out 16 months after GPT-2 and cost a couple million dollars of processing power. If this accelerated rate of Moore’s law holds up, GPT-4 is just around the corner and will be incredibly powerful and relatively affordable. GPT-3 is already blowing people’s minds, so it’s difficult to fathom what GPT-4 will be capable of.
上面提到的人工智能语言软件GPT-3的功能是其前身GPT-2的117倍。 GPT-3在GPT-2之后的16个月问世，处理能力耗资几百万美元。 如果摩尔定律的这种加速发展趋势得以维持，那么GPT-4指日可待，它将变得异常强大且相对负担得起。 GPT-3已经引起了人们的注意 ，因此很难理解GPT-4的功能。
Exponential growth is difficult for the human mind to contemplate and to use for prediction. It would have been difficult for someone in 2000 or even 2010 to imagine our technological capabilities today. There’s no way Gordon Moore could have comprehended today’s technology even if you were there telling him about it. But the relentless march of his law has brought us to today.
指数增长很难被人的思维所思索和用于预测。 对于2000年甚至2010年的某人来说，要想像出我们今天的技术能力将非常困难。 即使您在那里告诉戈登·摩尔，他也无法理解当今的技术。 但是他的法律无情地推动着我们发展到今天。
Now it appears that his law has accelerated. That means that the future is going to start coming even faster and that the developments that occur between now and 2030 or 2040 are going to alter society in some pretty fundamental ways. If you thought smartphones changed things significantly, you ain’t seen nothing yet.
现在看来他的法律已经加速了。 这意味着未来将开始更快地发展，从现在到2030年或2040年之间发生的发展将以某种非常根本的方式改变社会。 如果您认为智能手机已经发生了重大变化，那么您什么也没看到。
总结过去展望未来
