Shouldn’t We Really Be Talking About Augmenting Intelligence with AI?
Artificial Intelligence. AI. Somehow, when reading those terms, you sometimes feel a slight chill passing down your spine. The cinema has been playing on our fears of self-thinking machines and autonomous computers for almost 100 years. And Elon Musk, whose Tesla, Inc is developing its own AI chips, has even claimed that “AI is far more dangerous than nukes.”
Obviously, massive progress has been achieved in the past decades with AI enabling us to create music, pictures and even solve health and farming challenges that conventional approaches could never have efficiently resolved. So, it can’t be all that bad. Because of the negative connotations of the terminology, some people, such as Nick Ismail, are starting to recommend that AI should actually be referred to as ‘augmented intelligence’.
Today’s AI solutions are typically very good at efficiently tackling a single task. This is why a face recognition system can’t recognize music, and an AI music identification app can’t be used to analyze health DNA samples. What AI does well, is recognizing trends and outliers in masses of pre-prepared data in a manner that the human brain finds exceptionally challenging. Thus, humans analyze an issue, collect and sort data that highlights it, and then run that data through an AI program to get answers. The AI is only really doing the number crunching.
When considered from the perspective of augmenting human intelligence, it is clear that this term better reflects what AI actually achieves for us today. It helps us be more efficient and enhances our intelligence. It can also remove human bias and emotion, which can skew our ability to make an appropriate choice, especially when under pressure.
In a project for a global pharmaceutical company, an AI helped to improve patient safety by optimizing the locations teams should be working at. But, in order to achieve that result, it required teams of human data analysis employees to sort through the data to analyze and select the optimal sources. This included financial information, emails, and site visit reports. In such an example, AI is obviously doing some serious work but not doing something that humans couldn’t do at all. And, without the human support, it couldn’t have undertaken the task at all.
Experts also agree that human-like artificial intelligence is still a long way off. A survey of experts by the MIT Technology Review estimated that AI will require another 15 years to imitate a human retail salesperson, while human labor automation still lies around 120 years off.
AI, thought of in terms of artificial intelligence, only serves to inflate expectations of its capability, and imply a risk that currently isn’t there. Perhaps it is better to consider AI as an augmented intelligence tool to solve, in the words of Ada Lovelace, “that which human brains find it difficult or impossible to work out unerringly”.