AI – Between mass empowerment and mass confusion
What comes to your mind when you think of “Artificial Intelligence”? Do you think of robotics? Autonomous cars? Autonomous warehouses? Self-improving algorithms? The extinction of humankind? Whatever your perception of AI is, you’re probably thinking either too narrowly or too broadly about it. Both is dangerous. You might think that AI is merely the next buzzword invented by the big tech companies to sell their products and services. Or you may think AI will take over the world, replace humans, and dominate life on earth. Whatever your perception of AI is – maybe you´re inspired and confused at the same time – the machine age of marketing has arrived. We talk to Alexa to add items to our shopping cart; we ask Google to direct us to the next sushi restaurant in a city where we’ve never been; in just the click of a mouse, the cryptic symbols of a foreign language miraculously become legible.
AI – Back to the future
If we want to understand the role and impact that AI has on business and society, we have to take a brief look back in time. As novel as AI sounds, it is not new. The term itself was coined in 1956 in a proposal by an elite group of computer scientists and mathematicians who organized a summer workshop called the “Dartmouth Conference.” One of the opening paragraphs in the original proposal envisioned a future where “machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves.” When you read this, you might assume it’s from one of the latest AI conferences in the Silicon Valley or elsewhere. The truth is, AI has been around for decades and researchers have been working to a large extent on the same problems, from computer vision to understanding natural language. So what happened? The field of AI cooled off between the 1970s and late 1980s: today that time is known as the “AI winter.” Governments significantly reduced the funding of research programs and corporations lost faith in the strong claims made in the early days of AI. The reasons for the slowdown are connected to a number of factors, but computing power and the ability to process large quantities of data was a critical, limiting factor.
AI – Reloaded
A new era of AI research started in the late 1990s; IBM’s Deep Blue became the first computer that was able to beat chess grandmaster Garry Kasparov in 1997. At the same time, major research institutions around the world – and particularly the Japanese government – started investing heavily in the development of a new generation of computing systems. Meanwhile, the dot-com bubble attracted hitherto unseen amounts of seed funding for tech companies; data storage prices started an unseen decline; and computing power increased exponentially. Despite a short dip with the burst of the dot-com bubble, the technical infrastructure and developments in the machine learning community paved the way for many of the devices and services we take for granted today. In short, these developments which leveraged the use of AI led to the market domination we see today by global companies like Amazon, Google, Alibaba and Baidu.
AI – Solving well-defined problems better than humans
But what exactly is AI, and how does it affect our lives? AI can be classified into two broad categories, Artificial Narrow Intelligence (ANI) and Artificial General Intelligence (AGI). ANI captures the ability of machines to solve problems with respect to a narrowly-defined, specific goal. Think of your phone, for example. Have you ever wondered how it’s possible for all these pictures of your spouse or the people around you to be grouped together? How can your phone “know” these people, and seemingly classify and group these individuals accurately? Your phone – or more precisely, the software that runs the processing of your pictures – doesn’t know these people. Your phone also doesn’t care about these people. The software on your phone is conducting a very specific and well-defined task: To find pictures with the same or similar-looking people. These classification tasks are all made possible through AI. Whether classifying emails as spam or classifying the people you’re with in your pictures, tracking your location data to make specific restaurant suggestions, or simply transforming the input of your voice when you talk to Siri or Google Assistant into machine-readable text output: In all these cases, AI is used to solve a well-defined task without human intervention.
Do you remember the first time IBM Watson won the Jeopardy quiz show against Ken Jennings, the top Jeopardy champion? The system’s intelligence was rather limited: the system was “simply” able to look up the answer in a gigantic database faster than humans. AI has already outperformed humans on a plethora of tasks, from identifying melanoma better than doctors to knowing your psychological profile better than your closest friends, merely by analyzing your Facebook Likes. All these tasks are characterized by the solving of well-defined problems, and the majority of applications we see today are reflections of these narrow forms of intelligence or ANI.