Instead of saving costs by replacing humans with bots, Brynjolfsson notes, growth increases people’s productivity. Even better, some of the economic value of that productivity accrues to workers because their increased labor becomes more valuable. It’s not all fueled by tech’s billionaire owners.
The catch is that growth is difficult. When you’re just mimicking human behavior, you’ll know if you’ve nailed it (more or less). (A computer can play checkers: win!) But inventing a form of AI that works differently from the way humans operate requires more imagination. You have to think about how to create hand-fitting silicon superpowers with abilities unique to people, like our vague, “aha” intuition; Our common sense is logical; And our ability to creatively deal with rare, edge cases.
“Look at what’s there and say, ‘Okay, can we substitute a machine or a human in there?’ It’s 100 times easier to think. The really hard thing is to say, ‘Let’s imagine something that’s never existed before,'” Brynjolfsson said. “But ultimately that second way is where more value comes from.”
At the Stanford Institute for Human-Centered AI, director Fei-Fei Li wanted to find out what people actually want automated to do. Her team went to the US government’s “American Time Use Survey,” which details people’s daily tasks. Li’s team selected 2,000 everyday tasks that could feasibly be performed by AI and robots, asking people to rate how much they would like the task automated, with “zero Hell no, I don’t want robots doing itand maximum organism Please, I’m dying to make this a robot” Lee said.
“Open a Christmas present for me” is zero; “Toilet cleaning” is high. Clear enough, but in the middle are tricky things like “recommend a book”. The only way to find out what people want is by asking them—not by building AI based on science fiction fantasies.
Here’s another bug: It’s not always clear how the two types of AI differ.
DALL-E and other image generators can be argued to be a pure Turing play because they reflect the human ability to create art. The internet is currently groaning under the weight of articles claiming that human artisans are going to be made serially unemployed by AI. But creators can use apps to punch above their weight, for example, when a video game designer uses Midjourney to create art for a space shooter. It looks like augmentation.
Furthermore, most jobs are difficult completely Automate more than you think. In 2016, deep-learning pioneer Geoff Hinton argued that we should stop training radiologists because “in five years, it’s absolutely clear that deep learning is going to outperform radiologists.” (That could take 10 years, he added.) But there are still tons of radiologists working, and likely in the future, because the radiologist job is more complex than Hinton suggests, says Andrew McAfee, a colleague and co-author of Brynjolfsson’s, who co-directs the MIT Initiative on the Digital Economy. AI may be better at spotting potential tumors in scans, but that’s only a small part of a radiologist’s job. The rest involves preparing treatment plans and interacting with fearful patients. Tumor-spotting AIs could look to augment those doctors.
To steer companies away from Turingism, Brynjolfsson suggested some changes to government policy. One area ripe for reform is the US tax code. Currently, it taxes labor more heavily than capital, according to recent work by the Brookings Institute. Companies get better tax treatment when they buy robots or software to replace humans due to write-offs like capital depreciation. So the tax code essentially encourages firms to automate workers off the payroll, instead of keeping them and raising them.
“We subsidize capital and we tax labor,” Brynjolfsson said. “So right now we have entrepreneurs—whether they want to or not—trying to figure out ways to replace human labor. If we reverse that, or just level level playing field, then entrepreneurs will find a better way. That could be a way out of the trap.