3 Levers People Pull to Solve Problems AI Can’t
3 Levers People Pull to Solve Problems AI Can’t
What Still Separates Humans from Robots. Reviewing ‘Framers’ by Kenneth Cukier, Viktor Mayer-Schönberger, and Francis de Véricourt
Americans experience between 2.8 and 3 million infections by drug-resistant germs every year, resulting in 35,000 to 48,000 deaths, depending on how and what you count. Health officials and others have warned about the risks of so-called superbugs for some time. But what’s the right response?
One answer is to develop better antibiotics. Easier said than done. “Conventional drug development mostly focuses on finding substances with molecular ‘fingerprints’ similar to ones that work,” explain Kenneth Cukier, Viktor Mayer-Schönberger, and Francis de Véricourt in their book, Framers: Human Advantage in an Age of Technology and Turmoil.
Alas, superbugs snicker at such attempts.
“New antibiotics are so close in structure to existing ones that bacteria quickly develop resistance to them, too,” the authors explain. So what can you do when the very molecular properties that guide drug developers toward effective treatments are more easily evaded by superbugs?
Enter MIT professor of artificial intelligence Regina Barzilay. Instead of finding new drug candidates by looking for those easily-evaded similarities, Barzilay and her team built an algorithm to scour more than a hundred million different molecules looking for any with ample antimicrobial properties—the more surprising and dissimilar from known compounds the better.
After years of work, Barzilay and her team found their germ killer, one they never would have discovered following the old methodology. But as singular as this breakthrough was, the authors tell Barzilay’s story because it represents a capacity we all have and employ on a daily basis: the ability to frame and reframe problems in search of more effective solutions.
Building a Frame
Whenever humans face a situation, we approach it with a frame—that is, a mental model of the situation. Frames represent our conceptions of what we see. Sometimes these constructions come to us instantaneously, served up by our subconscious. Other times they come after long periods of analytical engagement. Either way, our frames are how we conceive situations, their possible trajectories, and our potential responses.
Cukier, Mayer-Schönberger, and Véricourt identify three primary factors in any given frame:
causality
counterfactuals
constraints
Start with the first, causality. Our brains are wired to assume cause-and-effect relationships between concepts: this causes that; one thing leads to another; a particular action produces a certain result. Of course, humans are also expert at mistaking correlation for causation. “If you’re headed to the grave,” as Jack White sang, “you don’t blame the hearse.”
By counterfactuals the authors mean imagining the situation differently than it appears. The ability to entertain alternate realities is how we plan, test theories, and develop strategies. Because we can imagine the world as it both is and isn’t, we have more agency to act within it. If you don’t like Option A’s outcome, what about B, C, or D? You choose.
Finally, constraints. A good frame captures whatever relevant details we require for an effective response. We don’t treat cancer with hammers or fill empty gas tanks with chicken stock. Constraints allow us to apply causal reasoning and entertain counterfactuals but still keep our response within useful bounds.
If we think of these three factors as variables, we discover that most improvements in the human condition come from adjusting or replacing one or more of them.
Not a subscriber? Take a moment and sign up. It’s free, and I’ll send you my favorite quotes on books and reading.
Resetting the Frame
For causality, when we learn A is related to B in a certain way, we can affect one by adjusting the other. Principles, rules, and laws assume as much. In the hard or soft sciences the “lawishness” of an observation depends on how reliable or consistent that relationship appears.
We can go wrong of course—when, for instance, we misapply a causal understanding from one setting to another. Journalist Michael Blastland highlights this problem in The Hidden Half. But the ability to generalize from specific instances and create broadly applicable rules is one thing that makes us human. So far neither horses nor robots have come up with Newton’s laws of motion, the principle of noncontradiction, or the biscuit method.
Similarly, humans have the unique ability to posit endless counterfactuals. Certain animals do show abilities with counterfactuals; at least we can surmise as much. But no species can do it to the extent humans can. This includes our ability to guess what others think and guess what they think we think—the ability of Person A to imagine what Person B imagines what Persons C, D, and E imagine, and so on. What researchers call theory of mind is just counterfactuals at work or the gym, on Zoom or Twitter, at church or in the car line.
Changing the counterfactuals—posing another possible scenario—allows us to play with cause and effect, to pull different levers and see alternative paths of action. Cukier, Mayer-Schönberger, and Véricourt provide several examples. Elon Musk, for instance, wanted his company, SpaceX, to develop and build reusable rockets—a nut NASA never cracked.
Musk figured it out.
Most every great business or technology story involves a person, usually a team, changing the counterfactuals. Doing so enables them to do what Musk did next: He changed the constraints.
Frames suggest their own limits but also reveal where the limits of another frame might not apply. In NASA’s case, they couldn’t imagine getting a rocket back to earth for reuse without wings, which created difficulties they couldn’t overcome. Meanwhile, Musk said, Who needs wings? We can use rocket thrust instead. He changed the variables, reset the frame, and landed a rocket for reuse just like he imagined he could.
The Human Advantage
When Barzilay’s algorithm unsurfaced her germ killer, the media jumped on the AI angle. Newspaper headlines gave AI the credit. “But that missed the real story,” say Cukier, Mayer-Schönberger, and Véricourt.
The ability to frame and reframe is something unique to humans. Artificial intelligence can’t do it, argue the authors, anymore than manatees can.
It wasn’t a victory for artificial intelligence but a success of human cognition: the ability to rise up to a critical challenge by conceiving it in a certain way, altering aspects of it, which open up new paths to a solution. Credit does not go to a new technology but to a human ability.
Humans first framed the problem one way, then reframed it another. They built the algorithm, trained it what to look for, and then determined the search field it worked in. The computer did the grunt work, but humans did what only humans can do (at least for now): combine and recombine the variables of causality, counterfactuals, and constraints to arrive at a successful solution.
This insight seems worth remembering as we engage with the recent developments of OpenAI (another company cofounded by Musk) such as ChatGPT and related applications. And it’s helpful when observing how useful AI can be when implementing our reframes, as it did with Barzilay’s algorithm.
Of course, there’s a caution buried in this dependence on humans. An advantage can become a liability if we allow some of our lesser impulses to intrude. When we are unwilling to consider new frames for reasons of ego, self-protectiveness, tribalism, and so on, we lose potential answers to our problems and risk doubling down on counterproductive solutions—another observation that can help us process the implications of AI.
No one human—or group of us—has all the answers to the challenges or opportunities we face. We need each other, including the friction of competing frames, to arrive at workable answers and breakthrough solutions. That’s true for the workplace, marriages, religion, politics, really any setting where people do anything meaningful.
“We may not know the solutions to the problems before us,” say Cukier, Mayer-Schönberger, and Véricourt, “but we know how to go about finding them. . . . The answer is to embrace what humans do well, to cherish our unique cognitive capabilities, and to refocus our minds on our ability to frame.”
Thanks for reading! If you enjoyed this review, please share it with a friend.
More remarkable reading is on its way soon. Subscribe today so you don’t miss out!