Marc Benioff: We're on the cusp of an AI revolution

Over the last 30 years, consumers have reaped the benefits of dramatic technological advances. In many countries, most people now have in their pockets a personal computer more powerful than the mainframes of the 1980s. The Atari 800XL computer that I developed games on when I was in high school was powered by a microprocessor with 3,500 transistors; the computer running on my iPhone today has two billion transistors. Back then, a gigabyte of storage cost $100,000 and was the size of a refrigerator; today it’s basically free and is measured in millimeters.

Even with these massive gains, we can expect still faster progress as the entire planet – people and things – becomes connected. Already, five billion people have access to a mobile device, and more than three billion people can access the Internet. In the coming years, 50 billion things – from light bulbs to refrigerators, roads, clothing, and more – will be connected to the Internet as well.

Every generation or so, emerging technologies converge, and something revolutionary occurs. For example, a maturing Internet, affordable bandwidth and file-compression, and Apple’s iconic iPhone enabled companies such as Uber, Airbnb, YouTube, Facebook, and Twitter to redefine the mobile-customer experience.

Now we are on the cusp of another major convergence: big data, machine learning, and increased computing power will soon make artificial intelligence, or AI, ubiquitous.

AI follows Albert Einstein’s dictum that genius renders simplicity from complexity. So, as the world itself becomes more complex, AI will become the defining technology of the twenty-first century, just as the microprocessor was in the twentieth century.

Consumers already encounter AI on a daily basis. Google uses machine learning to autocomplete search queries and often accurately predicts what someone is looking for. Facebook and Amazon use predictive algorithms to make recommendations based on a user’s reading or purchasing history. AI is the central component in self-driving cars – which can now avoid collisions and traffic congestion – and in game-playing systems like Google DeepMind’s AlphaGo, a computer that beat South Korean Go master Lee Sedol in a five-game match earlier this year.

Given AI’s wide applications, all companies today face an imperative to integrate it into their products and services; otherwise, they will not be able to compete with companies that are using data-collection networks to improve customer experiences and inform business decisions. The next generation of consumers will have grown up with digital technologies and will expect companies to anticipate their needs and provide instant, personalized responses to any query.

So far, AI has been too costly or complex for many businesses to make optimal use of it. It can be difficult to integrate into a business’s existing operations, and historically it has required highly skilled data scientists. As a result, many businesses still make important decisions based on instinct instead of information.

This will change in the next few years, as AI becomes more pervasive, potentially making every company and every employee smarter, faster, and more productive. Machine learning algorithms can analyze billions of signals to route customer service calls automatically to the most appropriate agent or determine which customers are most likely to purchase a particular product.

And AI’s applications extend beyond online retail: Brick-and-mortar stores still account for 90% of retail sales, according to the consultancy A.T. Kearney. Soon, when customers enter a physical store, they will be greeted by interactive chat-bots that can recommend products based on shopping history, offer special discounts, and handle customer-service issues.

Advances in so-called “deep learning,” a branch of AI modeled after the brain’s neural network, could enable intelligent digital assistants to help plan vacations with the acumen of a human assistant, or determine consumer sentiments toward a particular brand, based on millions of signals from social networks and other data sources. In health care, deep-learning algorithms could help doctors identify cancer-cell types or intracranial abnormalities from anywhere in the world in real time.

To deploy AI effectively, companies will need to keep privacy and security in mind. Because AI is fueled by data, the more data the machine gains about an individual, the better it can predict their needs and act on their behalf. But, of course, that massive flow of personal data could be appropriated in ways that breach trust. Companies will have to be transparent about how they use people’s personal data. AI can also detect and defend against digital security breaches, and will play a critical role in protecting user privacy and building trust.

As in past periods of economic transformation, AI will unleash new levels of productivity, augment our personal and professional lives, and pose existential questions about the age-old relationship between man and machine. It will disrupt industries and dislocate workers as it automates more tasks. But just as the Internet did 20 years ago, AI will also improve existing jobs and spawn new ones. We should expect this and adapt accordingly by providing training for the jobs of tomorrow, as well as safety nets for those who fall behind.

AI is still a long way from surpassing human intelligence. It has been 60 years since John McCarthy, a computer scientist and nominal father of AI, first introduced the term during a conference at Dartmouth College, and computers have only recently been able to detect cats in YouTube videos or determine the best route to the airport.

We can count on technological innovation to continue at an even more rapid pace than in previous generations. AI will become like electrical current – invisible and augmenting almost every part of our lives. Thirty years from now, we will wonder how we ever got along without our seemingly telepathic digital assistants, just as today it’s already hard to imagine going more than a few minutes without checking the 1980s mainframe in one’s pocket.