Podcast: The next step in AI is training machines to think “slower”
Nokia Bell Labs believes understanding human behavior is the key to developing disruptive innovations in artificial intelligence and machine learning.
In his book Thinking Fast and Slow, Nobel Prize winning psychologist Daniel Kahneman posited that human beings process information in two different ways. The first mode, called System 1, is fast, unconscious and energy efficient. The second, System 2, is the type of deliberate problem-solving thinking that allows humans to address difficult tasks. So, while System 1 thinking is what causes us to immediately pull back our hands when they encounter something hot, System 2 thinking allows us to solve puzzles, navigate our cars through dense traffic or reason out complex arguments.
Kahneman may have developed his fast-vs.-slow model as a means for understanding human behavior, but it’s also proving to be an invaluable tool in Nokia Bell Labs’ artificial intelligence and machine learning research, which is the subject of the latest edition of the Future Human Podcast. This week, Nokia Bell Labs’ Sean Kennedy, leader of the Algorithms, Analytics, Augmented Intelligence & Devices Lab, describes how Nokia Bell Labs is exploring the next frontiers in machine learning by training machines to solve System 2 problems.
Machines have already mastered many of the System 1 problems we’ve thrown at them. Take image recognition: machines can identify objects in images exponentially faster than any human. But a problem that machines can’t easily tackle is inferring from the information presented in a still image what is about to happen next. The goal of Nokia Bell Labs’ machine learning research is to not only train machines to solve these more complex System 2 problems by thinking in the same deliberate ways a human would, but also to execute such “slow” thinking as quickly as they would System 1 solutions. According to Kennedy, cracking the ceiling of System 2 machine learning will unlock a previously unimagined level of productivity.
Such advances in artificial intelligence will be integral to a future in which human augmentation technologies will boost our cognitive abilities. But AI and machine learning technologies are also poised to make an immediate impact today. Over the coming months, we will be highlighting how Nokia Bell Labs AI innovations are being applied to current problems, from helping track the spread of the Covid-19 pandemic to enabling more intelligent networks that can head off problems before they arise.
