Deepfake detection algorithms will never be enough

Spotting fakes is just the start of a much bigger battle

Illustration by Alex Castro / The Verge

You may have seen news stories last week about researchers developing tools that can detect deepfakes with greater than 90 percent accuracy. It’s comforting to think that with research like this, the harm caused by AI-generated fakes will be limited. Simply run your content through a deepfake detector and bang, the misinformation is gone!

But software that can spot AI-manipulated videos will only ever provide a partial fix to this problem, say experts. As with computer viruses or biological weapons, the threat from deepfakes is now a permanent feature on the landscape. And although it’s arguable whether or not deepfakes are a huge danger from a political perspective, they’re certainly damaging the lives of women here and now through the spread of fake nudes and pornography.

Hao Li, an expert in computer vision and associate professor at the University of Southern California, tells The Verge that any deepfake detector is only going to work for a short while. In fact, he says, “at some point it’s likely that it’s not going to be possible to detect [AI fakes] at all. So a different type of approach is going to need to be put in place to resolve this.”

SPOTTING DEEPFAKES IS LIKE SPOTTING VIRUSES: AN EVER-CHANGING CHALLENGE

Li should know — he’s part of the team that helped design one of those recent deepfake detectors. He and his colleagues built an algorithm capable of spotting AI edits of videos of famous politicians like Donald Trump and and Elizabeth Warren, by tracking small facial movements unique to each individual.

These markers are known as “soft biometrics” and are too subtle for AI to currently mimic. These include how Trump purses his lips before answering a question, or how Warren raises her eyebrows to emphasize a point. The algorithm learns to spot these movements by studying past footage of individuals, and the result is a tool that’s at least 92 percent accurate at spotting several different types of deepfakes.

Li, though, says it won’t be long until the work is useless. As he and his colleagues outlined in their paper, deepfake technology is developing with a virus / anti-virus dynamic.

One deepfake detection algorithm works by tracking subtle movements in the target’s face.

Take blinking. Back in June 2018, researchers found that because deepfake systems weren’t trained on footage of people with their eyes closed, the videos they produced featured unnatural blinking patterns. AI clones didn’t blink frequently enough or, sometimes, didn’t blink at all — characteristics that could be easily spotted with a simple algorithm.

But what happened next was somewhat predictable. “Shortly after this forensic technique was made public, the next generation of synthesis techniques incorporated blinking into their systems,” wrote Li and his colleagues. In other words: bye bye, deepfake detector.

Ironically, this back and forth mimics the technology at the heart of deepfakes: the generative adversarial network, or GAN. This is a type of machine learning system comprised of two neural networks, operating in concert. One network generates the fake and the other tries to detect it, with the content bouncing back and forth, and improving with each volley. This dynamic is replicated in the wider research landscape, where each new deepfake detection paper gives the deepfake makers a new challenge to overcome.

Delip Rao, VP of research at the AI Foundation, agrees that the challenge is far greater than simple detection, and says that these papers need to be put in perspective.

One deepfake detection algorithm unveiled last week boasted 97 percent accuracy, for example, but as Rao notes, that 3 percent could still be damaging when thinking at the scale of internet platforms. “Say Facebook deploys that [algorithm] and assuming Facebook gets around 350 million images a day, that’s a LOT of misidentified images,” says Rao. “With every false positive from the model, you are compromising the trust of the users.”

SOCIAL MEDIA PLATFORMS LIKE FACEBOOK DON’T HAVE A CLEAR DEEPFAKE POLICY

It’s incredibly important we develop technology that can spot fakes, says Rao, but the bigger challenge is making these methods useful. Social platforms still haven’t clearly defined their policies on deepfakes, as Facebook’s tussle with a fake Mark Zuckerberg video recently showed, and an outright ban would be unwise.

“At the minimum, the videos should be labeled if something is detected as being manipulated, based on automated systems,” says Li. He says it’s only a matter of time, though, before the fakes are undetectable. “Videos are just pixels, ultimately.”

Rao and his colleagues at the AI Foundation are working on approaches that incorporate human judgement, but others argue that verifying real videos and images should be the starting point, rather than spotting fakes. To that end, they’ve developed programs that can automatically watermark and identify images taken on cameras, while others have suggested using blockchain technology to verify content from trusted sources.

None of these techniques will “solve” the problem, though; not while the internet exists in its current form. As we’ve seen with fake news, just because a piece of content can be easily debunked doesn’t it mean it won’t be clicked and read and shared online.

More than anything else, the dynamics that define the web — frictionless sharing and the monetization of attention — mean that deepfakes will always find an audience.

Take the recent news of a developer creating an app that lets anyone make fake nudes of photographs of clothed women. The resulting images are obviously fabrications. When you look at them closely the skin and flesh is blurred and indistinct. But they’re convincing enough at a glance, and sometimes that’s all that’s needed.

Data Vision: This X-Ray App Could Help Hospitals Cut Costs

artificial intelligence x-ray app

Hospitals can feel pain, too, and computer scientist Karley Yoder is using artificial intelligence to treat it.

Yoder and her colleagues at GE Healthcare are working on apps that enable healthcare providers to analyze the reams of data their facilities generate, quickly diagnose problems and propose treatment. “Hospital operators know in their gut about problems, but they can’t quantify them,” Yoder says. “We give them the tools to identify and resolve those pain points.”

Yoder is spending this week in Chicago at the annual meeting of the Radiological Society of North America (RSNA), the world’s largest gathering for radiologists and other medical professionals. She and her team arrived with a new app that could make chest X-ray imaging more efficient.

X-rays represent more than half of all medical scans, and chest X-rays account for 50 percent of that volume.

Chest X-rays also leave a lot of room for error, Yoder says. She and her team have been searching for the root causes that can lead to bad scans and require the re-scanning of patients. “If you are a clinician taking an X-ray and your patient moves or you have a setting on your machine that inadvertently overexposes the image, you have to repeat the X-ray,” she says. “This exposes the patient to more radiation, slows down the patient flow and also adds extra paperwork because hospitals in the U.S. must report these instances to regulators.” These reports can take a day or more to compile, she says.

The app Yoder and her team developed takes care of the reporting automatically and also allows hospital managers to quickly spot problems. “It automatically pulls all of the information about rejected and repeated X-ray images into a one-page dashboard on the desktop that’s easy to review,” she says. “Administrators can drill in and see if there is a problem with a specific machine or a technician maybe needs more training. Once you quantify and identify the problem, you can quickly deploy a solution and move on.”

GE has been in the medical imaging business for more than a century. Above is an antique X-ray machine inside a GE Healthcare complex in Buc, France. Top: X-rays represent more than half of all medical scans. Images credit: GE Healthcare.

Yoder works in San Ramon, California, GE’s digital headquarters, but writing smart code is only one part of her job. The other is to make sure that healthcare practitioners won’t even notice the application running on their machines. “It’s like Netflix,” she says. “If you had to go to another application or open a new screen to get your movie recommendations, I doubt you’d use it.” She says that “there is always skepticism with new technology. We have to deliver it in a seamless way into the existing workflow.” She says that several GE Healthcare X-ray machines can already run the app and “the list is growing.”

The first iteration of the app will help hospitals reduce patient exposure to radiation, keep bureaucracy at bay, save time and make their radiology departments more efficient. But future iterations of the app or other similar advanced analytics technology may give technicians suggestions and maybe even help doctors with diagnoses. “When I think about deep learning in the healthcare space, I really think about it in three steps,” Yoder says. “We need to build the deep learning model and make sure it does what we want it to do. Next, we need to turn it into a product and deploy it. The third part is where the AI works and it gets really exciting because it allows me to work with the data and refine the model.”

Yoder says that one day artificial intelligence technology could help doctors send patients to the right specialists faster. It could also help them prioritize patients. “The machine can send them an alert: ‘Hey, take a look at this,’ ” she says. “I think that we are going to start seeing systems like this in many places in the healthcare field.”

And beyond, as well. Yoder’s colleagues in San Ramon work on smart software for aviation, energy and other industries where GE is active. She recently discussed her work with an engineer who suggested a possible use of the app in the oil-and-gas field, which is using industrial-grade X-ray machines to image parts for pipelines and other components. “We are very early in these conversations, but it’s the same technology,” she says. “But if you can scan hundreds of images and use an app to identify conditions like pneumothorax, you can probably use it to find a problem with your pipeline.”

Uli Ap Artificial Intelligence of Virtual Reality

Artificial Intelligence of Virtual Reality is a survey of Uli Ap’s site-specific immersive installations disorienting the visitor in a cross-reality zone between virtual, physical, and mental realms. Selected installations deal with issues of sacred and transitional spaces that generate sound through immediate interaction. Enveloped by soundscapes of Sacred Danger and That Side Where Real Is, you enter the altar of Anti-Gravity Reality: Inside the Black Cube or Brain AI transporting you into a transdisciplinary encounter between Virtual Reality, Artificial Intelligence, Robotics, and Neuroscience. There, you donate your emotions to a machine. Sounds at interventions to provoke your anxiety are the ones of Sacred Danger and That Side Where Real Is. All installations are connected. And standing in the middle of the ineradicably vast space of Ambika P3 you hear them merging with a sound of your heartbeat.

Installations:
Anti-Gravity Reality: Inside the Black Cube or Brain AI [prototype] Sacred Danger [Part II] [the Kim Fielding Award for experimental arts] That Side Where Real Is [Crypt@St.Mark’s St Mark’s Church Kennington] Red Film & The cReature Film

ANTI-GRAVITY REALITY: INSIDE THE BLACK CUBE OR BRAIN AI is an interactive cross-
reality installation that collides Virtual (VR), Physical, and Mental Realities.

Human emotions and feelings are being consumed by a living AI organism in order to create itself in both virtual and physical realms. The system measures your anxiety level and feeds itself with it through biofeedback sensors. Biosignals from electroencephalogram (EEG), galvanic skin response (GSR), and pulse sensor manipulate the environment. Virtual Reality acquires Physical qualities and tactile sensations. The experience is totally disorienting. Interacting with structures produced by their own brain, a participant unknowingly becomes a performer. Each experience will be unique and the space constructed at its culmination will be catalogued and transformed into an outside anti-gravitational pavilion. Recorded, it will travel to a gallery again as 360 VR installation.

The project merges art, science, and technology and stands for erasing labels, identities, disciplines, and geographical borders by advocating freedom of choice and movement. At no given moment, a participant is sure whether they are still in virtual or already in physical space. Or might be somewhere in space of their own imagination. It is a transdisciplinary encounter between Virtual Reality, Artificial Intelligence, Neuroscience, and Robotics.

SACRED DANGER [the Kim Fielding Award for experimental arts]

Part I: outside installation Part II: multi-channel sound-video installation transferring experience of a physically built space through the digital realms

Sacred Danger – immersive site-specific outside installation or a structural intervention in a landscape reinventing a sacred space. A project is inspired by my research trip to Blaenau Ffesteniog in North Wales, where I have been taken by local people to abandoned slate queries. By experiencing a sensation of the innermost fear and danger during this journey – those slate queries could collapse any time – I have compared them to Gothic cathedrals, that one of a beauty and attractive seduction though simultaneously the one of a danger and manipulation.

Visible from far away fields, Sacred Danger sits on a sharply inclined hill like a weightless cloud inviting you to enter into its inside. Although, the very moment you find yourself there, a situation is reversed into insecurity and vulnerability by exposing your body to the attack of a fluctuating under the strong wind structure built out of just 6mm plexiglass penetrated by massive 1inch steel rods. Dragon shape of a zig-zag foundation is confusing in its refractive magnifying reflections, and willing to escape to a stunning sea side horizon you are bounded to a disappointment – you are inside the space with no exit. The exit is blocked by a transparent plexiglass which obviously you do not see unless by touching it on approaching. In effect, you would try to exit Sacred Danger on every turn of a dragon only to realise that transparency is an illusion and all your senses and perceptions are disoriented. Chameleonic nature of Sacred Danger, with a great sensitivity reacting on each of the atmospheric conditions, makes the
structure change its appearance within nearly every hour. Transparent throughout the day. Frosted in the early morning by dense fog. Turning flaming red in the sunset. With sliding raindrops in the evening. The structure claims its own presence to be unbelievable. A space with no exit dissolves in nature by confronting it. Triggers the mind by challenging and provoking it. Delusional crystal clear transparency of plexiglass transforms Sacred Danger into a totally invisible volume suspending threaded bars in the air. It is deliberately created to transfer the feeling of insecurity, instability and danger – you just simply don’t believe that the structure holds itself up – and this is exactly how it should be. Though, each experience would be totally different, somehow you are supposed to be scared inside – and so you do.
On entering Sacred Danger through a hole in a bush, you are signing an agreement with nature by excepting a possibility of the existence of something that logically could not have ever existed.

THAT SIDE WHERE REAL IS

9 site-specific immersive Installations challenging our perceptions of reality Crypt@St.Mark’s Kennington, London, UK [TSWRI] Nine Site-Specific Immersive Installations Questioning Perceptions of Material Reality
& Dream-Like States of Consciousness

Components of ‘That Side Where Real Is’:
Cracking incorporating The cReature film [Room1] Transparent Invisibility [Room2] Suspended Dreams to Eat the Time [Room3] There Is No One In Here [Room4] Vertical Immersion [Room5] All How It Was [Open Room] Black Penetration Into And Through Horizontal Dimensions [Corridor. Part I] Black Penetration Into And Through Horizontal Dimensions [Corridor. Part II] Red Gloss One String [Entry Hall]

Multi-component installation ‘That Side Where Real Is’ – is a series of physically constructed, site-specific, inter-connected installations challenging our perception of space and the idea of the space as such built from ephemeral materials or materials that become ephemeral when being constructed into the space. To go through the installation is mentally, intellectually and emotionally intense experience activated by constantly changing hallucinational spaces on the way you move. At the installation context, a body becomes a tool penetrating a space [human element penetration] confronting a built environment battling with the dissolution and vagueness.

The exhibition’s title, ‘That Side Where Real Is’ refers both to the visual challenges conveyed through the materials used and the way they trick the mind by disorienting the body, making it to squeeze in-between something that seems to be impossible to enter, perceive something that human eyes cannot simply see unless they are positioned at certain angles or fixed at the points of the space where the light is reflected. Emotionally and intellectually challenging chameleonic geometry that constantly changes into hallucinational spaces while your body is moving. It is a division between physical and abstract, rational and intuitive – with a big question mark on which of them is real.
A peculiarity of the site is echoed in here – being in an underground crypt, empty and dark, with the whole universe of life happening on the upper layer inside the building of the church itself; and, then, beyond the point of the church – in a real life on the streets. Due to the very nature of the crypt ‘That Side Where Real Is’ deals with a notion of a death/life duality. Sound at the installation is derived from the material elements themselves – be it cracking dry leaves or transparent fishing wire strings, or black elastic shock cords. By reacting on a bodily movement, singular elements generate noises as a response. Those noises are important part of the perception. ‘That Side Where Real Is’ continues to experiment with light, its effect on visibility and depth, and a coloured light as a separate material. The cReature film is as process-based as the overall construction and being organically buried inside dry cracking leaves narratively tells a birth-life-death sequence with the absolute awareness of its senselessness. Being divided on two sides a real has a character of a vague and obscure presence that is constantly transferred in-between. A space of the film prolongs disorienting irony of transparent and invisible strings, deepening it down towards the very extremes.

Uli Ap Artificial Intelligence of Virtual Reality Ambika P3 University of Westminster
35 Marylebone Road London NW1 5LS [entrance through the main reception]

About The Artist

Uli Ap is a trans-disciplinary artist researching into a cross-reality zone between virtual, physical and mental spaces where definitions between the mediums and disciplines are no longer possible. Provocatively colliding art, science, technology, architecture, and journalism, the artist creates intellectually challenging and bodily disorienting site-specific immersive interactive installations to question our perceptions of the self and various levels of reality.

Uli lives and works in London and travels the world extensively to exhibit and take part in art residencies. Ap undertook artist’s residencies in the British School at Rome, ‘International
Artist-in-Residence’ in Aso, Minamioguni-machi, Japan, received inaugural Kim Fielding Award for experimental arts, Cardiff, Wales and Emerging Artists Promotions Grant, London, UK. Ambika P3 exhibition gives a comprehensive insight into the artist’s practice. Uli Ap’s work is included at the forthcoming “Weather the Weather” a SciArt Initiative exhibition at the New York Hall of Science.

Ambika P3 is a 14,000 square ft space developed from the vast former concrete construction hall for the University of Westminster’s School of Engineering. Built in the 1960s, it is a distinctive space for contemporary art and architecture which presents a public programme of solo and group exhibitions, education projects and talks and events involving the academic community from the Faculty of Architecture and Built Environment and Westminster School ofMedia, Arts and Design.

Mobile Technology and Home Broadband 2019

37% of Americans now go online mostly using a smartphone, and these devices are increasingly cited as a reason for not having a high-speed internet connection at home

As the share of Americans who say they own a smartphone has increased dramatically over the past decade – from 35% in 2011 to 81% in 2019 – a new Pew Research Center survey finds that the way many people choose to go online is markedly different than in previous years.

Today, 37% of U.S. adults say they mostly use a smartphone when accessing the internet. This share has nearly doubled since 2013, when the Center last asked this question. At that point, 19% of Americans named their smartphone as their primary device for going online.1

Younger adults are especially likely to reach for their phones when going online. Fully 58% of 18-to 29-year-olds say they mostly go online through a smartphone, up from 41% in 2013. Still, this growth is evident across all age groups. For example, the share of adults ages 30 to 49 who say they mostly use a smartphone to go online has nearly doubled – from 24% in 2013 to 47% today.

These trends are part of a broader shift toward mobile technology that has changed the way people do everything from getting news to applying for jobs.

Indeed, mobile devices are not simply being used more often to go online – some Americans are forgoing traditional broadband at home altogether in favor of their smartphone. A majority of adults say they subscribe to home broadband, but about one-in-four (27%) do not. And growing shares of these non-adopters cite their mobile phone as a reason for not subscribing to these services.

Among non-broadband users, 45% say they do not have broadband at home because their smartphone lets them do everything they need to do online, up from 27% in 2015. At the same time, the share of non-broadband users who say their smartphone is the most important reason for not having a high-speed internet connection where they live has nearly doubled over the same time period (from 12% to 23%). And while affordability remains a commonly cited barrier, the share of non-broadband adopters who say the cost of a monthly subscription is the most important reason for not having these services has fallen from 33% in 2015 to 21% today.

In addition, 80% of these non-broadband users say they are not interested in getting high-speed connections at home.

These patterns underscore the reliance that a minority of Americans have on their smartphone for internet access. Some 17% of U.S. adults are “smartphone-only internet users” – meaning they report owning a smartphone but do not have a traditional high-speed internet connection where they live. This share has roughly doubled since 2013, when 8% of adults fell into this category.

These are some of the key findings from Pew Research Center’s telephone survey of 1,502 U.S. adults conducted Jan. 8 – Feb. 7, 2019.

A majority of Americans subscribe to high-speed internet service at home or own a smartphone, but digital gaps remain between some groups

Fully 81% of adults now say they own a smartphone, up slightly from 77% in 2018. Smartphone ownership is relatively common among Americans of different economic, educational and racial and ethnic backgrounds. Adults ages 65 and older are the only major demographic group in which a clear majority do not own this type of device, but ownership rates vary even within this population: 59% of those ages 65 to 74 are smartphone owners, but that share falls to 40% among those 75 and older.

The share of Americans who say they have broadband service at home has also risen slightly over the past year – from 65% in early 2018 to 73% today. But compared with smartphone ownership, there are more pronounced variations in broadband adoption across demographic groups. For example, 92% of adults from households earning $75,000 or more a year say they have broadband internet at home, but that share falls to 56% among those whose annual household income falls below $30,000. That 36-point gap in broadband adoption between the highest- and lowest-income groups is substantially larger than the 24-point gap in smartphone ownership between these groups. Educational differences follow a nearly identical pattern.

And as has been true with other Center surveys, there are double-digit gaps in home broadband adoption by community type, as well as by race and ethnicity.

Roughly one-in-four lower-income adults are “smartphone only” internet users

For some Americans without traditional broadband, smartphones may help bridge these gaps in connectivity. Overall, 17% of Americans are now “smartphone only” internet users – meaning they own a smartphone but do not subscribe to broadband internet service at home. This share has roughly doubled since 2013. At that time, just 8% of Americans relied on their smartphone for internet use at home. The 2019 figure is, however, somewhat lower than the percentage who said this in 2018 (20%).

In general, smartphone reliance tends to be more common among groups who also have lower levels of broadband adoption. Some 26% of adults who have a high school education or less are smartphone only internet users. By comparison, 16% of those with some college experience and only 4% of college graduates fall into this category. Lower-income adults are also more likely than those in higher-earning households to be smartphone only internet users.

Racial and ethnic gaps are also present: One-quarter of Hispanics and a comparable share of blacks are smartphone only internet users, compared with about one-in-ten whites.

Smartphone users are now more likely to go online using their mobile phone than with some other type of device

For those who own a smartphone, these devices now outpace more traditional means of accessing the web. Some 46% of smartphone owners say when using the internet, they mostly do so on their phone. This represents a double-digit increase from 2013, when 34% of these users said this.

At the same time, the share of smartphone users who say a desktop, laptop or tablet computer is their primary device for going online has fallen from 53% in 2013 to 30% today. Roughly one-quarter of smartphone users say they equally use a cellphone or a desktop, laptop or tablet computer when going online, up from 12% six years prior.

Smartphone users’ preferred method of going online varies significantly by age. While 60% of smartphone owners ages 18 to 29 and roughly half of those ages 30 to 49 say they mostly use their cellphone to use the internet, those shares fall to about one-third or less among those ages 50 and older. On the other hand, smartphone users ages 50 and over are about twice as likely as those under the age of 50 to say they mostly access the internet through a computer of some kind (43% vs. 21%).

Beyond age, there are also educational differences in the way smartphone users access the internet. Roughly six-in-ten smartphone owners with a high school diploma or less (57%) say they mostly use the internet via a cellphone, compared with 45% of smartphone users with some college experience and 34% of those with a bachelor’s degree or more. By contrast, smartphone owners who are college graduates are about twice as likely as those with a high school education or less to say they mostly use a computer to go online (40% vs. 21%). Similar gaps are present by household income and race and ethnicity. (For more details on how different demographic groups access the internet, see Appendix A.)

A growing share of non-broadband users cite their smartphone’s capabilities as the reason they do not subscribe to high-speed internet service at home

The reasons non-broadband users cite for not having broadband service have shifted in recent years. Affordability remains a commonly cited barrier to subscribing to these services, but the shares of non-broadband adopters who mention cost has declined since 2015.

Half of non-broadband users today say they do not subscribe to broadband because the cost of a monthly subscription is too expensive, while 31% say the cost of a computer is too expensive. In 2015, those shares were 59% and 45%, respectively..

Meanwhile, a growing share of non-broadband users credit their smartphone as a reason why they forgo a subscription to high-speed home internet service. Some 45% of non-broadband adopters say they do not have high-speed internet at home because their smartphone lets them do everything online that they need to do. This represents an 18 percentage point increase from 2015, when only 27% of non-adopters cited their smartphone as a reason for not having home broadband.

Roughly four-in-ten non-adopters report having other options for internet access outside of their home, while 22% say broadband service not being available or available at an unacceptable speed where they live is a reason for not subscribing to home broadband. Both of these shares have changed little in the past four years.

When asked to name the most important reason they do not subscribe to home broadband, more non-adopters point to the functionality of their smartphones today than in the past.

Some 23% of non-broadband users say the most important reason for them not having a broadband connection at home is because their smartphones let them do everything they need to do online – roughly doubling the share (12%) who said this in 2015.

And while financial factors remain a frequently cited reason for non-adoption, fewer non-broadband users cite cost as the primary reason they do not subscribe to these services. Some 21% of non-broadband users say the monthly cost of service is the single most important reason why they do not have broadband at home, with an additional 6% attributing their non-adoption to computers being too expensive. In each case, both of these shares were higher in 2015 (33% and 10%, respectively).

And for non-broadband users who own a smartphone, the functionality of their smartphone is the single most important reason they cite for not subscribing to high-speed internet at home. Some 37% of non-broadband adopters who have a smartphone say this, statistically unchanged from 2015. At the same time, roughly one-quarter of non-broadband users who have a smartphone say either the monthly cost of a broadband subscription (22%) or the cost of a computer (4%) is the primary reason they do not have home broadband, down from 38% in 2015.

A majority of non-broadband users have never had high-speed internet at home and relatively few are interested in having it in the future

In previous Center surveys, Americans of all kinds have linked not having broadband to a number of disadvantages – including difficulties finding job information or less access to government services. Despite these perceived challenges, non-adopters show little interest in becoming home broadband users in the future.

Six-in-ten non-broadband users say they have neverhad high-speed internet service at home in the past, while one-third indicate that they had previously subscribed to these services. And when asked if they are interested in having home broadband in the future, most non-adopters are unenthusiastic about the prospect. Fully 80% of non-broadband users say they would not be interested in having broadband at home, while 18% think this is something they would consider in the future. These sentiments are on par with those found in the Center’s 2015 survey.