AI researchers estimate there’s a 50 percent chance AI will outperform humans at all tasks by 2060.
By 1997, computers were better than humans at chess. By 10 years later, they were better at driving cars than the average teenager, and they were better at playing Chinese game Go, rated 300 times harder than chess. But what if by 2060, artificial intelligence proved better at absolutely everything?
That was the takeaway from a recent study by Yale and Oxford universities. They polled 352 researchers in AI who determined there’s a 50 percent chance AI will be able to outperform humans in all tasks by 2060.
Health care, law, finance — these top minds predict that every industry you can imagine could be completely transformed in just a few decades. Does that mean enterprises will evolve out of needing any human employees in just 40 or so years?
Probably not, but leading industry executives working in AI imagine this new future will definitely feature a lot of intelligent automation. And instead of leaving humans in the dust, it will make us even better at what we do too.
“Every time there’s an advancement, we are using that as a ladder to our next step of evolution as human beings,” says Jay Klein, chief technology officer at Voyager Labs. “If you look back at the Industrial Revolution, the world didn’t stop after some inventions were being made. On the contrary, we continued to evolve.”
Ali Ghodsi, CEO and co-founder of Databricks, a San Francisco company led by the team that created Apache Spark and that provides a unified analytics platform for data analytics at scale, sees AI enabling humans but also doesn’t foresee a completely automated future with the human out of the loop.
“The way AI is being built is simply not at all in any way the way true human intelligence works, but it can augment us [in some domains] and do a much better job. But humans will still be super critical by 2060.”
AI has already gotten extremely good at pattern prediction, enabling so much of the big data revolution that’s going on in the industry right now. This fuels many other areas where the technology is currently outpacing humans, like image and video recognition. And it’s making great gains in areas such as natural language recognition.
Ghodsi says this AI prowess all boils down to excelling at solving defined problems in a highly defined structure. But outside of this structure is where AI will continue to need human influence.
“AlphaGo is now beating the human Go players in Go, and that’s awesome,” he says. “But ask the computer to reflect on its victory, and it has no clue what that means. If I asked a human that question, they would have an answer, and the computer would be clueless.”
Humans are still masters of tasks that involve creativity, emotional intelligence or formulating a problem in the first place, instead of just solving one. Take, for instance, programming itself.
“In the job of programming, there are almost no advances there,” says Ghodsi. “If it’s an open-ended problem that doesn’t have a clear, well-defined structure, then [AI] won’t be able to do it.”
For AI to excel outside its structured fields, it’s going to have to rely on human intelligence in many other areas. The common thread of all the big AI companies right now — like Google, Facebook and Microsoft — is they happen to have the world’s largest data sets, says Ghodsi. But to apply this sophisticated AI to another field, like medicine, the algorithm will need access to data sets that may not yet exist — like billions of X-rays of tumors, for example. And programmers won’t know how to let AI interpret that data without tapping into the decades of knowledge from top oncologists.
Klein agrees that AI must evolve, like all technology, to suit the needs of users.
“AI will follow the structural pattern of all innovation,” he says. “Consider the late ’70s — people were buying personal computers, but they lacked targeted benefits for consumers and businesses. They spent $4,000 for a screen and keyboard with limited usability. Remember how long it took for friendly word processors or tools like PowerPoint to emerge? Similarly AI is currently too generic to be meaningfully adapted for different industries. A significant amount of algorithmic development must be done for these vertical needs to be met. It will happen, but will take longer than people realize.”
So it seems it’s not time for everyone to permanently retire and let the robots do their jobs just yet. Humans will play an increasingly important role going forward, playing to their own strengths with AI as a powerful backbone. Businesses can get themselves ready for this AI revolution by properly building and managing their data, leveraging the expertise of their industry leaders and blending all that knowledge with artificial intelligence experts. Expect to see these huge changes to occur first in industries like health care, customer experience and transportation, where enterprises have spent the last decade refining their big data efforts.
Disclosure: Both Databricks and Voyager Labs are Merritt Group clients.