TED 2018: Thought-Reading Machines and the Death of Love
Ludwig Wittgenstein once imagined that everyone had a box with something in it called a “beetle.” Denying the possibility of private language, the philosopher wrote, “No one can look into anyone else’s box, and everyone says he knows what a beetle is only by looking at his beetle.” Wittgenstein meant that we learn a word by observing the rules governing its use, but no one sees another person’s beetle: “It would be quite possible for everyone to have something different in his box,” or nothing at all. An apparently intractable fact of life is that our thoughts are inaccessible to one another. Our skulls are like space helmets; we are trapped in our heads, unable to convey the quiddity of our sensations.
But how much longer will our thoughts be truly private? Mary Lou Jepsen, the founder of Openwater, wants to show me the beetle inside your box, and you the beetle inside mine. The San Francisco startup is developing an optical imaging system—sufficiently compact to fit inside a skull cap, wand, or bandage—that scatters and captures near-infrared light inside our bodies to create holograms that reveal our occluded selves. The devices could diagnose cancers as well as cardiovascular or other diseases. But because the wavelength of near-infrared light is smaller than a micron, smaller than the smallest neuron, Jepsen believes the resolution of the technology is fine enough to make thoughts visible too.
Jason Pontin (@jason_pontin) is an Ideas contributor for WIRED. He is a senior partner at Flagship Pioneering, a firm in Boston that creates, builds, and funds companies that solve problems in health, food, and sustainability. From 2004 to 2017, he was the editor in chief and publisher of MIT Technology Review. Before that he was the editor of Red Herring, a business magazine that was popular during the dot-com boom. Pontin does not write about Flagship’s portfolio companies nor about their competitors.
Jepsen unveiled Openwater’s technology at the TED Conference in Vancouver last week. Given the company’s much-publicized ambitions, attendees expressed interest and skepticism as they waited to enter the theater. On the darkened stage Jepsen showed how bodies are translucent to red light, popping a light-emitting diode in her mouth so that her head glowed from within like a bloody skull. She shone a beam through a cranial fragment, demonstrating that red light could penetrate even the thickest bone. She scattered light into raw chicken, making a hidden tumor visible, and then hid the tumor by doubling the quantity of flesh, illustrating how de-scattering would be necessary to see anything deeply buried in our bodies. Finally, she scattered light into a milky box with the optical properties of a brain, focused light within the box, and created a hologram that visualized an object just a few microns wide. Although the components onstage were bulky lab equipment, Jepsen promised Openwater would refashion integrated circuits—using liquid crystal displays as light emitters and camera chips as sensors—to make holographic devices as cheap and easy to use as our smartphones. “Developer kits in a year, commercial products the next,” she swore. When Jepsen finished, she smiled, relieved that the demo gods had been kind, and the crowd stood and cheered.
A few days earlier, in a backroom where Openwater’s chief engineer fussed over his lights and lenses, Jepsen explained that the company’s promise depended on combining these elements: proof of the entire body’s translucence; holographic techniques, some dating to the 1960s; and Asian silicon manufacturing, which can make new chip architectures into commercial products. Openwater may be less than two years old, but Jepsen has been thinking about a holographic scanner for decades. She is uniquely suited to the challenge. Her early research was in holography; she led display development at Intel, Google X, and Facebook Oculus; and she has shipped billions of dollars of chips. She also has personal reasons to care about brain tumors: 20 years ago, she endured months of sickness before an MRI revealed a tumor that was later removed. (I’ve been friendly with Mary Lou Jepsen for many years.)
There would be uncontroversial benefits to Openwater’s technology. According to Jepsen, two-thirds of humanity have no access to medical imaging. Magnetic Resonating Imaging (MRI) machines occupy an entire room and cost several millions of dollars, plus half a million a year to maintain; in the poor and developing world, MRI machines are simply unavailable. Openwater might license its technology to medical device companies, which would manufacture the products and pay for their regulatory approval, populating village hospitals, clinics in the rich world, and homes with inexpensive, wearable devices. The medical possibilities aren’t limited to reading bodies but extend to what Jepsen calls “writing,” too. Infectious agents can be killed by light, and light can make therapies more effective. “Because of the photodynamic quality of cells, you could cure all kinds of diseases or reduce chemotherapy doses by 25 times,” she speculates.
But, of course, what startles about Openwater is the proposal to read and write thoughts. The idea derives from Jack Gallant, a cognitive neuroscientist at UC Berkeley, who decoded movies shown to subjects in a functional MRI machine by scanning the oxygenated blood in their brains. The images Gallant recovered are blurry, because the resolution of fMRI is comparatively coarse. Holography would not only see blood better but capture the electrochemical pulses of the neurons themselves. Jepsen thinks in images, and has always been attracted by the arts; she imagines painters or musicians expressing themselves by thinking of pictures or sounds. “We’ve been constrained by how quickly we could talk or by the facility of our hands.” Pressed, she imagines more science fictional scenarios: creators collaborating by throwing each other thoughts, like children lobbing balls in play.
As a cheaper alternative to MRI, Openwater’s technology is ambitious. Rodney Brooks, founder of Rethink Robotics, warns, “There are lots of things in a series that all have to come together to make it all as expected.” The promise of brain scanning is even more uncertain. Ed Boyden, a neuroscientist at the MIT Media Lab and a pioneer in the field of optogenetics—where genetically modified neurons are controlled with light—cautions, “The underlying physics of overcoming the scattering of light in tissue is an interesting field with well-established results. But we don’t know how thoughts are computed by the brain. Scaling up the technology to the size of the human brain and proving that it can be applied in a safe way presents a great engineering and clinical challenge.”
Jack Gallant, who is working on a Facebook project where people would type by thinking of words, is more forgiving. He notes that everything we think, feel, or remember is represented in biophysical processes in the brain. “Any instrument that can measure any aspect of these processes will recover some decodable information about brain states. The issue is how much of the potentially decodable information is, in fact, decoded by a specific technology.” Because we don’t know how thoughts are computed, nor how much of that information would be decoded, the brain scanner might see images or hear music with wonderful clarity and yet be blind and deaf to vast tracts of human thought. What is the shape and color of pity? It is like something for a sentient being to feel qualia, and no one knows how a machine would convey such integrated experiences.
A cheaper alternative to MRI might be a moral imperative, but a brain scanner could be abused for purposes of surveillance and control. Vaguely, Jepsen says that anyone who donned her cap would consent to its operations. But it’s easy to recall how social media exploits our frailties and appetites to distract and madden us, even though we consent to its terms of service. A real machine-brain interface might be fantastically worse than Facebook. More terrifyingly, a brain-scanner could be an instrument of explicit coercion. Would interrogators and torturers use such devices on their subjects? Jepsen’s response to these humanist worries is to call for dialog. “When you look at where the technology is, I think it’s inevitable. But we’re not going to decide what’s ethical; we want to work with lots of different organizations to make those decisions. Ultimately, [conversation] should result in some kind of new declaration of human rights.”
Even if a brain scanner were possible and its uses benign, the creatures who enjoyed it often might not be recognizably human. Almost everything we traditionally value depends on the privacy of our thoughts. Parents lie to young children about their accomplishments; adult children deceive aging parents about their declines. Even when we are in love, our desire to commune with another is frustrated. We are most human when we are gentle, or when we strive to explain ourselves. But if our thoughts were always open to each other, our individual identities would begin to blur. We’d be different kinds of social beings, living under a different social contract.
Openwater’s name was suggested by the musician Peter Gabriel, a friend of Jepsen’s, who rhapsodized in an essay about the “transparent waters” we’ll swim in one day. “We will need to create ‘swimming lessons’ to teach us how to be comfortable being open, honest and exposed … ready to float and navigate in these waters of visible thought,” he wrote. I’m not ready. I want to stay on the shore, clutching my beetle (in the end, the only thing that is truly mine), vainly trying to convey the nacreous colors on its back.