Algorithms are increasingly affecting our lives . But most of the time it works, we don't notice , Only when the algorithm goes wrong can we notice its existence . Only then did we think back to the world of algorithms — The ones that dominate all the computers around us , Rules that are increasingly difficult to understand for humans , How dependent is it . Once the algorithm goes wrong , We will remember how fragile we are （ Think Skynet ）.
Pedro Domingos It took a lot of time to think about algorithms . His new book ,《The Master Algorithm: How the Quest for the
Ultimate Learning Machine Will Remake Our
World》 It's an introduction to that world , And report on the latest situation . He thinks that , We live in an age of algorithms , We're seeing them reshape our world in an unprecedented way .
What is the impact of algorithm on human beings ? What is the difference between human thinking and computer thinking ? What happens when the machine finally learns to learn everything ?Jesse Hicks Interview on these issues
Why are so many people not aware of algorithms in your so-called algorithmic era ? What is the machine learning mechanism behind it ?
Computer can't do anything without algorithm . Your cell phone , notebook , automobile , There are algorithms everywhere in houses and appliances . But the algorithm is invisible , You can only see the bright appearance , I can't see what's going on inside .Siri
Use algorithms to understand what you say ,Yelp Use algorithms to select restaurants for you , vehicle GPS
Use algorithms to find the best route for you , The reader uses algorithms to pay for you . Companies use algorithms to screen job seekers , Mutual funds use algorithms to trade stocks , Mobile phones use algorithms to mark suspicious calls .
“ normal ”
The difference between algorithm and learning algorithm is that , The former relies on manual programming by software engineers , Step by step, tell the computer what to do , The latter relies on reading the data ： Na , Here is the input , Here is the output we want , How can I turn one into another ? What's striking is that , From chess to medical analysis , The same machine algorithm can learn almost unlimited things — Just give it the right data .
In the title of the book “ Main algorithm master algorithm” What is it? ? It follows Ray Kurzweil What's the difference between the singularity of ? What progress can the main algorithm bring ?
The main algorithm is an algorithm that can learn anything from data . Give it the movement of the planet , Inclined plane , Pendulum data , It will discover Newton's law . Give it DNA
Crystal structure data, it can find the double helix . With the data on your smartphone, it can predict what you're going to do next and how to help you . You can even find a cure for cancer by learning a large database of cancer patients' medical records .
Algorithms may also bring us home robots ; use WWB（World Wide Brain, Vincristine ） replace WWW（ web ）, Answer your questions directly instead of showing you the web page ; as well as
360° Recommendation system for , Know not only you but also your best friend , Not only can I recommend books to you , film , And the date , work , Houses and tourist destinations are everything in your life .
Kurzweil The singularity of artificial intelligence is that artificial intelligence surpasses human intelligence , The moment that we couldn't understand . Or more precisely , It's singular “ Horizon event horizon
”, Just like the horizon of a black hole is the point where even light cannot escape . If there is no main algorithm , We won't get to the singularity so soon . With the main algorithm ,AI
Of course, it will speed up , But we can still understand a lot about the world , Because under our leadership AI
Still able to serve us . We may not know how they produce results , But we can know what these outputs will do for us , Or they won't be . in addition , There are things in the world that we can't understand . The difference is that , This incomprehensible part of the world today is made by ourselves , This is, of course, an improvement .
You said this field is currently “ tribe ”
The situation of separatism , Some machine learning algorithms perform better in solving specific problems , But no algorithm can beat any other algorithm ： There is a lack of a unified theory that can be applied to everything we know so far , A theory that can lay the foundation for the development of the next few decades or even hundreds of years . The assertion itself is grand . What is the rationality of the main algorithm ? Why now
“ tribe ” We can't unite yet ?
It can be proved mathematically , Even the simplest learning algorithm , Just give enough data , You can learn anything . therefore , There is no doubt that the main algorithm exists , And the researchers of each algorithmic tribe confirm that they have discovered it . But the key is that the algorithm must be able to use reasonable data and computation to learn what you want it to learn . We can give two empirical examples ： Nature offers at least two examples of how algorithms can learn anything ： Evolution and the brain . So the main algorithm exists , The question is whether we can find out exactly , Write it down in its entirety , Just as physicists express the laws of physics by formula （ It is also an algorithm in itself ）.
Unfortunately , Machine learning 5
A tribe is like a blind man and an elephant ： He who touches his nose thinks it's a snake , He who touches his foot thinks it's a tree , The one who touched the tooth thought it was a cow . We need to step back and look at the panorama , See how all these parts come together . Ironically , It may be easier for people who don't know what to do .
His book begins with a quotation Alfred North Whitehead The words of ：“ The progress of human civilization is achieved by increasing the number of important operations that can be performed without thinking .”
Whether this conclusion holds or not , but “ reflection ” Undoubtedly, it is closely related to civilization and human nature . Thinking is unique , Even decisive human activities . therefore Nicholas Carr
Others are against outsourcing thinking , Because it reduces our humanity — The fear is that the lack of thinking will lead us to be more robotic （ In a broad sense ）. meanwhile , We are worried again “ reflection ”
machine ： You mentioned the apocalyptic AI like Skynet . Does the computer have the ability
“ reflection ”? Or is it a unique human activity — If so , What are the differences between future human thinkers and machine learners ?
Famous computer scientist Edsger Dijkstra
Yes , Whether a computer can think about this is as important as whether a submarine can swim . The important thing is that computers can solve problems that human beings solve through thinking — And the scope of these problems is expanding . Through machine learning, computers have even solved problems that we don't know how to program them to solve — They came up with it on their own . So the line is very vague , And it's changing all the time .
I disagree Nicholas Carr The argument that outsourcing thinking will destroy human nature — contrary , It also enhances humanity , Because it allows us to think about better things . And that's exactly what it is Whitehead
What's the point . Socrates doesn't like writing , Because it makes people forget things . Fortunately Plato wrote down his ideas , So now humans can remember them . Writing enhances our memory ,Google
It has raised it to a higher level . It doesn't make us any more stupid , It's getting smarter .
The book ends with ,“ People worry that computers will become too smart to take over the world , But the real problem is that they are stupid and have taken over the world .” Can you explain what it means ?
Hawking and Elon Musk These celebrities expressed concerns about AI , It's a threat to human existence . but “ Skynet ” Such evil AI
The idea of taking over the world is far fetched . The problem is , People confuse intelligence with people . In Hollywood movies AI
And robots are always humanoid , But in reality, they are very different . Computers have no will of their own , Emotional or conscious . They're just an extension of us . As long as they solve the problems we set , As long as we set the boundary of the problem and examine the solution , Computers can be infinitely intelligent without threatening us .
But that's not to say you don't have to worry at all . Just like any other technology , People also use them for evil purposes AI. But the most important thing is ,AI
Will provide what we ask for, not what we really want , And that could lead to injury . Computers have made important decisions today — Who should get the job , Who should get credit , Who should be labeled as a potential terrorist . And they tend to make mistakes , Because they lack common sense . But the solution should be to make them smarter , Not more stupid . So what we should worry about is not
AI Too much , It's too little .