Hello World: Being Human in the Age of Algorithms
by Hannah Fry ***** Five of five stars
Algorithms control our lives. That's a given in this age of unlimited information. They help us find jobs, mates, cars, houses, and most everything else. More and more they are relied upon by doctors, educators employers, insurance companies and all of big business.
This book tackles the important topic of algorithms and compares them to the fragile and imperfect abilities of the humans they are supposed to serve. "Hello World" is a basic command that's given to first time programmers, and Hannah Fry takes us behind the scenes to tackle what the programmers are seeing.
There are two types of algorithms- the rule-based ones that we take for granted that are repeated instructions for getting from A to B, and machine-learning ones that have come to dominate our world, also known as artificial intelligence, or AI. The book covers six critical areas where AI is being utilized right now, and goes over in depth the pros and cons of their abilities.
1- Data is mined by new learning algorithms to figure out our behaviors, looking for patterns in our buying, voting and living habits. Some of the programs ares scarily accurate. Large numbers of data brokers now trade in our purchasing data and build profiles of our likes and dislikes, which drives which kind of advertising we can see. Fry looks at the Cambridge Analytica scandal and how the data company was able to identify vulnerable voters and push information their way, much of it fake. This kind of manipulation will almost certainly continue.
2- Fry looks at our justice system- specifically how judges decide on sentence length and whether to release accused criminals on bail. Databases of past results are now being used to score people who have been arrested and while the scores are helpful, they are also ridden with biases. Judges, unfortunately, are even worse, giving vastly different sentences for the same crime depending on their mood, the time of day, or their feelings toward the accused.
3- She then looks at the medical field, and points to new programs that can help spot cancerous cells on slides much more accurately than humans can. Computers are great at finding patterns, and can catch things humans can't especially if different humans deal with a patient over an extended period. There are attempts now to diagnose illnesses using computers, as well as projects to look at DNA from genetic providers like 23 and me. While the speed is impressive, mistakes can happen and privacy is a big concern.
4- Self driving cars are supposedly the next big thing for AI, and Fry looks at the challenges in this field. Humans make terrible mistakes when driving, and many of them can be disastrous. Machines can drive cars too, but there are many real-world issues that self-driving cars can't anticipate. Having a hybrid system where the car drives itself most of the time and relies on the human as backup looks to be a bad idea because humans are lazy and too trusting of computer prowess.
5- The chapter on crime covers how hot spots for criminals can be predicted by machines, something that police departments are utilizing more and more. Computers use pattern recognition to predict where crimes are most likely to happen, which allows officers to head off problems ahead of time. Crime shows on tv are all big on facial recognition technology, which depends on incredibly complex algorithms, but are far from perfect.
6- The final chapter, oddly enough, is about how AI can predict and create great art and music. AI programs have written pieces of music that have fooled audiences. They can also use audience preferences to tell if something is likely to be popular, though this process is still far from perfect. We rely on social proof through such algorithms as Rotten Tomatoes, Billboard, and bestseller lists to tell us if something is good or not. We also let places like Amazon and Netflix suggest tv shows and movies to us because they can see what we watch the most.
So what's the answer? Is AI coming to take our jobs and be our master? Not for a while, but it does in general do a better job at most tasks than we do. It's faster, more consistent, and less prone to cognitive bias than human beings. Ms. Fry tells the precautionary tale of Stanislov Petrov, the man who single-handedly saved the world from nuclear war. In 1983 Petrov was working for the Soviets when his computer system alerted him to an incoming US attack. Proper protocol would have been to believe the computer and pass the information to his bosses. But Petrov, in his wisdom, overruled the computer and judged it to be a false alarm, which it turned out to be.
AI performs better than humans, but it still far from perfect, and the last chapter of this book is a wonderful recipe for working alongside algorithms, not beneath them. They are too prone to bias, error, and incomplete judgements to ever be completely trusted. There will always be a place for humans to step in an correct them when needed, but the challenge will be where and when.
Comments