The Smart Glasses Contract

When smartphones first came out, nobody in the industry predicted how we’d end up using them.   Nokia commissioned some fascinating research on the use and psychology of photo-taking, which they published as an interesting book – Mobile Photo.  Its main findings were that most rolls of film had a Christmas tree at either end, from which they predicted that smartphones would mostly be used for photos of friends and pets, which would be shown on the phone.  This was back in 2000, when social media was in its infancy and mobile networks barely supported the transfer of grainy photos.

Nobody in the industry saw where things would go.  Or go wrong.  What was a novelty has changed people’s lives.  So far, it has been overwhelmingly consensual.  Individuals have the choice of what they capture and upload.  Most probably don’t fully understand the implications og having their images online, and some abuse it, but there is a fairly well defined contract.  With smart glasses, that is about to change.  To illustrate why, let me ask a simple question.  Do you appear in any online porn videos?

I suspect everyone believes they can answer that question unequivocally.  A few years ago, you could.  Today, you can’t be quite so sure.  To explain why, let’s look at an everyday couple – Taylor and Pat.  They’ve just finished watching the latest series of Rivals and thought it would be fun to place themselves in their favourite sex scenes.  Their AI app is happy to oblige.  They upload the series clips along with links to lots of photos of themselves and enjoy watching the resulting  video. 

Human nature being what it is, a few weeks later they start thinking how it might look if they were taller, blonder, fitter or better endowed and go back to their AI to ask it to update the clips.  This time, the AI is likely to trawl the web for images which meet Pat and Taylor’s requirements, melding them into the bodies it has been asked to update, because that’s easier than transforming the images of Taylor and Pat from scratch.  Nobody will know who the new images are based on or where they came from, but if your image met Pat and Taylor’s requirements, it could be you.

As neither Pat nor Taylor are identifiable as the participants in their latest clips, they don’t feel quite so discreet about them and share them with a few close friends.  Once again, human nature asserts itself – some of their friends share them with their friends and eventually they end up on the net.  The people portrayed will probably never know, unless someone sees the clip and tells them.

We have no idea how much this happens, but we do know that it does.  There is a steady stream of court cases of malicious deepfakes being uncovered, with legislation being proposed to try to control them.  But the everyday use of your image is still largely uncharted.  Once you upload it, it’s essentially fair game for any AI model that is scraping the web for content.  At the corporate level, publishers and film studio are fighting the AI companies in the courts, claiming that training a Large Language Model (LLM) on copyright content is a breach.  In general, the US Government appears to support the AI companies’ approach, which is “just do it” – not least because the Government fears any restraint might jeopardise what they believe is the US’ “global leadership in artificial intelligence”.  Or, to put it another way, your image is deemed collateral fodder for the good of the AI industry and making America great again.

If the service is free, you’re the product

There is an adage in the internet world, which is “If the service is free, you are the product”.  The implication is that if you use a free service, whether that’s email, social media, or any other app, then the service provider can use any data or images you’ve upload to “improve” their business models.  Initially that was metadata that they used to develop new services or sell to advertisers.  As AI has evolved over the last few years it’s needed more and more data to train algorithms.  Almost every company involved in AI development has started trawling the web and social media for everything that is written and every image that has been uploaded.  As almost all of the social media companies are heavily involved with AI, they’ve all been happy to let your data be hoovered up, whether that’s photos, videos or text.  All of the AI applications can then use it largely as they please, without asking you permission.  Which means your online presence may be wider than you think.  And not necessarily where you think it is.

Up until now, you had a choice, albeit a limited one.  If you wanted to keep control of your data or images, so that they weren’t accessible to programs scraping the net, and your image stayed private, you just stopped uploading anything.  It’s essentially security by obscurity.  Some information will always get out as you’re captured in friend’s or work photos, but it’s limited and probably won’t have much metadata tagging it.  It does mean you’ll be living in a world without social media – a choice few would make today.  But it was an option.

The Smart Glasses Contract

The arrival of smart glasses changes the equation.  As soon as someone is looking at you, capturing and uploading your image without your knowledge, you lose the ability to opt out.  A short video will contain enough data to make it moderately easy for an AI to identify you, and anything that is uploaded can be scraped by other AIs and used.  Public privacy will disappear.

There is already a concern about smart glasses being used to capture information which is deemed private.  Many theatres and cinemas are banning smart glasses to stop performances being pirated.  Courts are banning them, not least because of evidence that jurors may be using them not just to record trials, but to request legal advice on the case they’re involved in.  There’s also growing outrage about their use in secretly filming and uploading images of girls and women.  These are all cases of misuse where legislation could help, but they conceal the fact that the previous social contract has changed.  With the usage model promoted by Meta, nobody’s image is exempt.  Their philosophy is that we are all grist to the AI mill, allegedly for the greater good of mankind.

Where the debate gets difficult, is that smart glasses can have enormous benefit.  People with accessibility issues find them invaluable.  But this is where we hit the dilemma. To be useful, the AI driving them needs a large amount of data to understand the world around you.  Limiting the quantity of that data would make them less useful, and they would probably never be economically viable to develop for the people who most need them.  On the other hand, a laissez fair approach to uploading everything with no restrictions (à laMeta) could generate a backlash large enough to kill the market for everyone.

Some manufacturers are very aware of this.  Even, another smart glasses manufacturer, believes that smart glasses don’t need to record, and have just launched a pair without a camera, aiming more at productivity.  It’s a very different concept and illustrates that the market really doesn’t know where it’s going at the moment.  Does the public want glasses that are always-on cameras, or would they be satisfied with head-up displays that responded to their current context?  Whilst companies like Meta may argue they’re the same thing, the privacy approaches that result are diametrically opposite.

The speed of deployment and lack of perception of the issues is a large part of the problem.  It’s led to the widespread feeling that Smart Glasses are predominantly used by “manfluencers”, with the media dubbing them “Pervert Glasses”  For those who remember the outcry against Google Glass, you’ve not seen anything yet.  Meta is not helping the debate.  Business Insider reports that Meta’s platform head, Adam Mosseri, claims the company would ban any user uploading videos showing them harassing others.  But that only tackles their deliberate misuse, and is an after the event sop.  So is their claim that they’ll stop users obscuring the recording light with bits or black tape.  These statements show that they totally fail to understand the unintentional undermining of personal perception.  This isn’t about deliberate harassment – that will always happen and can (in theory) be tacked with legislation.  It’s about chipping away at your ownership of your identity, and moving us to a world where anyone can be deliberately or  accidentally portrayed as something that is not yourself.  What Meta fails to understand is the value of personal identity. 

Rupert Goodwins, writing on The Register, makes a nice analogy, which is to compare smart glasses with dogs.  Few people would trust an XL Bully with their children, but millions trust labradors.  Which is why we try to control XL Bullies.  It’s not difficult to guess who Rupert considers to be he XL Bully.

He’s not alone.  There’s a business aphorism from Mark Zuckerberg which is often quoted by tech startups – “Move fast and break things”.  More and more critics of Meta’s Smart Glasses policy feel that needs to be amended to “Move fast and break people”.  Because people are being broken.  Many have already discovered that their image has been used without their knowledge or consent.  All they did was to upload photos of themselves, which have been trawled by AI models as they were being trained.  Although everyone knew that putting their image in the cloud made it public, few thought through the consequences. 

The subtle and insidious difference that comes with smart glasses is that the option to opt out disappears, as anyone can upload what they see to the web.  In most cases, it’s unlikely to be done maliciously – you’ll just be there as someone records their holiday or asks their AI for directions, but once your image is captured, you’ve lost control of it and given it to any app that wants to use you.  That is new, and is something that needs to be discussed, as once that contract is normalised, there is no return.