In chapter 4 you will learn how to estimate and control the costs of implementing AI. There is absolutely no point in adopting a new technology, in investing in it, unless it will make your organization more profitable and/or cost-effective. Every change entails risk, just like sitting there and doing nothing! Finance and tech have to work together to achieve AI success. Government organizations may not feel pressure to generate profit, but they sure can benefit the taxpayers by providing more—and better—for less.
In chapter 5 Joshua will address the platform you need to govern your data in an AI system. Your organization’s data is one of your most valuable assets! You will learn the EDEN method to keep your garden green and growing. We’ll say it again: AI without good data is a waste of time and money.
Chapter 6 lays out case studies about AI and legal controls. The law is your friend, not your enemy. Law is messy and imprecise, but in democracies it is the best way to curb abuses and guarantee personal freedoms. This chapter gives directions on how best to govern AI systems across the economy for the good of the state and society.
The afterword provides you with a simple, clear, workable ethical framework that can be applied to almost any organization where people work with data and automated systems.
So, if you want to wait for a future of perfect freedom, no restraints, and universal prosperity, you could waste your time and money on singularity science fiction, but we sincerely advise against it. AI is a very real technological advancement, and once achieved, these never go away. The future of your business and of our shared democratic, capitalist system requires your full attention to AI right now.
1.
https://www.humanbrainproject.eu/en/.
Chapter 1
7 AI Myths, 7 AI Rules
The term artificial intelligence is an old fund-raising gimmick. John McCarthy coined the term in 1955 to apply for Rockefeller money to pay for a conference at Dartmouth about “automatic computers.” The goal of the conference was “to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves.”[1] Computers are as far as away from that as ever, but in some applications they can at least appear to come close.
Today more than one thousand vendor companies use the term “AI” to sell their services or to raise money from investors to cover their expenses. Whether they actually use AI algorithms is another matter. Many exaggerate what AI can do, and there’s nothing new in that. In the past sixty years, we have gone through cycles called “AI winters” when soaring promises returned failed deliverables, underwhelmed investors, and sank research funding. Depending on how you count, an AI winter happens roughly once per decade.
We find ourselves in another hype cycle once again. Claims about AI capabilities have spun out of control in media and advertising, leaving the tight discipline of computer science firmly on the ground. Joshua and I do not, however, predict another AI winter because of the real strides being made in computer-processing speed and storage, and the explosive growth of available data. There will be a major shakedown among vendors, but the tech is only getting better. (I will explain how it works in chapter 2.)
The first step in your successful, profitable adoption of AI tech in your organization is to clear away the myths that cloud the real picture. Below are seven that are repeated all too often. Let’s make short work of them.
Myth 1: AI Is a Robot
AI and robots are not the same. AI is a family of data analytics procedures or algorithms performed by software run on hardware. Robots are contraptions equipped with sensors that produce digitized data, a central processing unit for that data, mechanical parts to complete tasks, and a power supply to run on. The robot’s data processor may or may not use AI algorithms.
The confusion is understandable. In journalism, stories about AI usually feature a picture of a robot or a digitized, humanoid face, probably because images of software, such as computer code or lots of 1s and 0s, are dead boring for most people.
Robots have caught people’s attention for millennia. From ancient Egyptian records, there are stories of statues of gods that gave advice or moved. In ancient Greece, there was a tale about a man made of bronze named Talos who guarded the island of Crete. Statues that come to life by one means or another are almost stock characters in fantasy literature. Little children everywhere love to imagine that their dolls, stuffed animals, and action figures are alive and interact with them. Big people don’t seem to want to give that one up too easily.
AI is not bound to any certain device. Many AI applications, perhaps most, process data with no input or output from any moving mechanism. AI in sales and marketing, HR, finance, customer support, education, legal, and government usually involves no robots at all. (Chatbots are another matter.) AI software, however, can be used to control virtually any hardware component that relies on data processing—from autonomous cars to automated welding arms to swarms of flying weapons, or whatever.
Robots can use AI in order to react to stimuli detected by their sensors and to adjust their motions or behavior accordingly. But AI doesn’t need a robot to make it worth your investment.
Even if AI is used to determine the actions of some robots, it does not attain anything close to human intelligence inside them, no matter how lifelike they may look or act. When you see a robot like “Eric,” “Kuri,” or “Jibo,” think “technologically animated plastic doll,” not “human replacement.” Saudi Arabia’s granting citizenship to a robot named “Sophia” was a publicity stunt. No robot, with or without AI, can guide itself according to human concepts of justice, equality, fairness, and equity, or cope with human unpredictability. Managers of the Henn-na hotel in Japan staffed it with more than 240 robots but then got rid of half of them. In trying to respond to the widely varied needs of individual customers, the robots made too much work for the humans.
Yes, robots have been developed for sex, and the company that makes them opened a robot brothel near New York, but was stopped in Houston. We have no more to write about this subject, but we have to point out the obvious: No robot can love. Nor can any AI system. As much as some people may adore them, robots will never love them back. AI guru Kai-Fu Lee seems to have needed a life-threatening bout with cancer to realize this truth.[2]
Myth 2: AI Knows What It’s Doing
The other popular myth is that AI is a hazard because it “wants” something—that is, to replace humans. Very prominent US entrepreneurs such as Elon Musk have issued warnings along these lines. Futurists who predict a coming “superintelligence” warn that AI or machine intelligence will outstrip the human in due time, with dire consequences.[3] Once we figure out “general artificial intelligence,” others claim, it will then figure out that it does not need us. Because it has to be plugged in in order to function, will start to defend itself from humans, using every conceivable means to keep the electricity on.
An AI system doesn’t “want” anything. It lacks volition—a will. It is a mathematical object that works to attain the goals defined by its programmers.
AI performance at rule-bound games, such as chess, Go, Jeopardy, Dota 2, and other competitive eSports, depends entirely on the data sets, rules, and goals established by the programmers. The appropriate means to victory do not really matter as long as the rules allow them. In a boating game experiment, the AI was extensively trained in the program, and it proved victorious, but only by crashing its boat into the wall as many times as possible.
AI can “learn” the software, not the spirit of the game, or competition, or camaraderie. AI can play well enough alone, but its record for team playing is abysmal. Some observers of the AI vs. human Dota 2 video game showdown remarked that the AI character pulled moves