In the early years of AI, computer scientists believed that artificial intelligence would soon match “the general intelligence of an average human being” and maybe even replace subject matter experts, such as lawyers, doctors, and accountants. Decades later, we know that AI did not reach these lofty goals by the 1990s or even the new millennium. But does that mean that AI will never live up to expectations, as industry insider John Horgan argues in a recent article in Scientific American, “Will Artificial Intelligence Ever Live Up to Its Hype?”
Sure, AI still presents challenges that technologists of the 1960s, 70s, and 80s thought would be solved by now. For example, algorithms are only good at the tasks they’re trained to do. This problem is called “nonrecurrent engineering” and limits AI’s ability to mimic human intelligence. AI is also only as good as the data it was trained on. If that data was biased or otherwise tainted, the resulting algorithm will be as well.
Many AI projects operate in a “black box,” meaning that there is no way of knowing how the AI used its training to arrive at a conclusion. This problem is exacerbated by researchers in the field; only 15% of AI studies released in 2020 were open source. Perhaps some companies experimenting with AI are using this characteristic of AI to their advantage. Some reports of the benefits of new AI appear to be public relations stunts on closer inspection. DeepMind, an Alphabet company, announced in 2016 that a Google data center reduced its cooling bill by 40% using DeepMind AI. However, since then, the results have not been replicated in any other data centers.
Perhaps most challenging for AI are human characteristics that can’t be quantified, such as emotional intelligence, common sense, evolution, and the ability to detect correlation from causation.
AI is changing our personal and professional lives nonetheless. It turns out, completely replacing humans with artificial intelligence is not the best way to use AI, at least not right now. Even if (or when) technology does allow AI to encompass all aspects of the human experience, there are ethical questions we must answer before doing so. Instead, today’s augmented intelligence plays to the innate strengths of AI as well as humans. As retired professor of technology management Jeffrey Funk writes in IEEE Spectrum, “software that automates tasks normally carried out by white-collar workers is probably the most promising of the products and services that AI is being applied to.”
After all, “all evidence suggests that human and machine intelligence are radically different.” Rather than attempting to replicate human intelligence in machines, the best uses for AI harness machines’ strengths. Robots don’t get tired or hungry, and they don’t make so-called “human error.” Additionally, AI’s natural language processing abilities continue to improve thanks to increased investment by established companies and startups alike. This makes communication between humans and machines easier, potentially solving some of the challenges of AI.
As Funk puts it, it’s an AI evolution, not an AI revolution. While that might disappoint some AI scientists, it’s still enough to change industry, education, and everyday life for the better.