The following article discusses the themes explored in the documentary, which includes substance abuse, mental health, gun violence and suicide.

We all know, or knew, that guy. Not in your social circle, but known nonetheless; someone’s older brother, cousin or drinking buddy. Whenever they had a captive audience they tell you tales of their exploits when they’re not kicking around suburban Lowestoft. In between puffs of cigarette smoke and the cheapest whiskey available, they’ll say they tried to join the army, but the recruitment people told them they were just too brilliant to waste in an infantry unit. Or they are an off-duty bodyguard who was lying low because The Mafia was looking for them (don’t ask why, shut up). Or that they had just signed a contract to replace The Undertaker at The Wrestling™ and would be jetting off to the US in the near future. The intensity of their testimony may, for a brief second, sucker you in, but you’ll soon realize that these people are more Walter Mitty than Walter White. Now imagine what that guy would look like if they’d been handed $100 million, and you’ll get a fairly decent pen portrait of John McAfee in his later years.

Running with the Devil: The wild world of John McAfee is a new documentary, arriving on Netflix on August 24th. It harnesses footage from the lost, unreleased Vice documentary On The Run with John McAfee, as well as film McAfee commissioned himself. It attempts to chronicle the life of the antivirus software pioneer from when he was named as a person of interest following the death of his neighbor Gegory Faull in Belize, through to his death in 2021. McAfee would spend his last decade on the run from pursuers, both real and imaginary, become embroiled in a cryptocurrency scam, try to run for US president (twice) and loudly declare that he refused to pay his taxes, which attracted the attention of the IRS. Arrested in Spain on charges of tax evasion, he died by suicide in his prison cell.

Devil is broken into three rough parts, each told from the perspective of the people in McAfee’s orbit at the time. Part one focuses on then-Vice editor-in-chief Rocco Castoro and legendary photojournalist Robert King, who accompanied McAfee on his escape to Guatemala. Part Two covers McAfee’s backstory and his relationship with ghostwriter Alex Cody Foster, with whom he sat for a series of interviews. Part Three shows how McAfee would eventually reconnect with Robert King, and asked him to become his personal biographer as he sailed on his yacht, mostly around South America. The footage is interspersed with commentary from McAfee’s partners, as well as Foster, Castoro and King.

Something that’s clear from both the footage and the contributors is that McAfee was obsessed with truth, but not always as you or I would understand it. There are several times when he fixates upon his legacy, his reputation, his image, his story and how he would be perceived. And yet the story was malleable, the facts unclear, and his behavior erratic – while on the run, he would buy a disguise and then proudly tell everyone in the store his name, and pose for photographs. McAfee’s behavior mirrors the cult leader who’s gone all-in on the grand deception, both in his use of charm, and his propensity for violence. More than once he’s pictured or discussed pointing a gun at friends and allies for what feels like nothing more than the pleasure of being a bully, or at least to remind everyone who had the power.

If you’re looking for some sort of truth, or grand coherent narrative to help you grasp who John McAfee was, however, you won’t get it here. That’s not a criticism of the documentary – McAfee loved to hint about who he was without ever saying it out loud, and always muddying his own water. There are scenes where he implies he is responsible for the death of both his abusive father and Faull, but never to anyone’s satisfaction. But it’s similarly clear that much of his bravado disappears when he’s faced with real consequences for his actions. Much is made, too, of his substance abuse, which seems to have supercharged his paranoia and delusional thinking.

Netflix

Much of the footage shot by King is low-res, untreated first-person digital video, although there’s little shakycam here. It instantly dates the footage back to the start of the last decade, and sets the scene perfectly given the turn-of-the-millennium anxieties it creates. It works here, too, because it captures the unpleasant stale air in rooms that haven’t had their windows opened for too long. Rooms scattered with dirt and loose tobacco flakes, a half-empty whiskey bottle resting on its side next to some bath salts and a loaded handgun. It helps capture the smallness of the man in his decline, especially as he rages against not the dying of the light, but to the world’s seeming indifference. I imagine that anyone trying to dock a yacht in a foreign country with a cadre of automatic weapons and mercenaries on board would be greeted with a frosty reception from the local police. But, for McAfee, it’s all part of the grand conspiracy the world has contorted around him, and it’s sad. But you can’t feel too much sympathy for him given the trail of destruction left in his wake, and there’s little closure offered for his victims here.

If there’s one thing I wish the film did better, it’s helping the audience keep track of who, and where, everyone is at each point. I’m not always a fan of documentaries with hand-holding narrators, but this is the sort of film that really needs you to have Wikipedia to hand. That’s not to say it’s not worth watching, both if you knew of McAfee or if the original saga had passed you by. But if it lacks something, it’s enough of a sense of place and time to help you keep track of all of the things that McAfee was up to, and when.

It’s funny, several of my colleagues met with McAfee over the years – including this Engadget Show segment back in 2013. (Back then, McAfee said that he was parodying and leaning in to his insalubrious reputation while he made his viral videos. The documentary makes it clear that there was perhaps more truth than he was prepared to admit.) I’d even walked past McAfee several times at CES, often sitting alone in a sparsely-attended corner of one of the smaller show halls. I often wondered if I should go and speak to him, but there was something of the That Guy even when he was ostensibly on his best behavior. I could imagine him clamping his hand on my shoulder, fixing me with his dark eyes and spinning a fresh bewitching tale of mystery and intrigue, although as it turns out, the truth was probably wilder.

We’re living in an age of micromobility, where city dwellers are given more options to get around than ever before. Bike-sharing schemes, which quickly swelled to include e-scooters, are big deals in major areas where the commuting distances are just that bit too long to walk. But scooters have yet to achieve the same legitimacy as the bicycle with mainstream users. In part, this is because it’s still a relatively emerging technology, but also because the scooter has the whiff of a toy about it. After all, the Segway was meant to revolutionize transportation, as it headed on its one-way journey toward the novelty store. Similarly, hoverboards never had any real shot at making its way across the rubicon that separates useless from utility. E-scooters are already on the other side of the water, but will they stay there long term?

It’s something that the team at Bo Mobility believes is entirely possible, so long as there’s some serious, grown-up effort put in today. The UK-based company was founded by Oscar Morgan, Harry Wills and Luke Robus, who met while working at Williams Advanced Engineering, part of the F1 team. They’ve spent the last three years working on Bo M, a serious e-scooter that’s been conceived from the ground up to be exactly that. It’s promised to be better designed, run better, and have a longer lifespan than any of the quasi-toy products that litter our streets today. I’ve ridden the prototype (two, in fact) and I already feel confident that it’s a quantum leap compared to what we currently have.

Daniel Cooper

The first thing you’ll notice about Bo is its shape, with a swan neck that arches gracefully from the steering column to the deck. There’s no hinge or fixing joint between the two, as it’s not built to fold or compress down in any way – you park it in its upright position. Ditching the fold means that you can massively increase strength, and gives its creators room to add in some extra features that you won’t find on your typical $399 scooter. “People say ‘you guys made a pretty scooter,’” says Oscar Morgan, co-founder, “but it’s really important to appreciate that it has a fundamentally different architecture … When you move from a tubular frame to a proper monocoque construction, you move all the stresses. It makes it stronger, but it also means that we can start to package stuff.”

And the most important thing in the package – nestled inside that beefy curved cowl, is a product called Safesteer. It is, for now, a series of prototypes, each being refined before the scooter’s launch next year. And it’s top secret, beyond the fact that it’s a hybrid analog / digital device that’s designed to improve balance and maneuverability. (What I have been able to get out of Morgan: It’s not a gyroscope, which was how something like the Segway managed to stay upright.) After all, most e-scooters require a little bit of a learning curve, especially the cheaply-made ones. It’s not for nothing that most e-scooters require you to keep both hands on the bars, unlike a bike which can be ridden one (or, if you’re brave, no) handed.

Daniel Cooper

“Going from a big bike wheel to a [small] scooter wheel,” said Morgan, “you lose all of your gyroscopic stability.” Safesteer will hopefully redress the balance, enabling you to ride around with the same level of indifference as you would with a bike. And, similarly, Morgan and Wills’ team has sweated the details in terms of weight distribution, wheel dynamics and suspension. Or, in this case, a lack of it. The pair found that replacing the 8-inch tyres seen on most scooters with bigger, 10-inch models offers far greater balance and grip. That enabled them to leave out the mechanical suspension, massively reducing weight, which improved maneuverability. You may think that the ride quality suffers, but the company is working on AirDeck, adding an elastomer to the footplate which will act as a shock absorber, similar to that you’d find on a pricey running shoe. Then again, you’re not going to be riding your e-scooter into too many potholes no matter how comfortable the ride is (and Bo’s is nicer and smoother than some I’ve tried).

And then there’s the load hook – which is such a small addition but one that the team, again, sweated over. Morgan explained that as part of the idea to pull people out of their cars and onto scooters, that they’d need some secure way to hold luggage. “You see lots of people riding with bags,” said Wills, “hanging them from the handlebars, and it's really dangerous when you’re riding around.” “[The load] unbalances and swings, and then you get a bit of tank slap when you hit a bend or corner,” he added. Instead, a centrally-mounted pair of hooks buried inside the monocoque will electrically roll out, letting you hook your bag onto the scooter’s center of weight.

Daniel Cooper

Naturally, users will be paying for the privilege of riding Bo when it makes it debut in the spring of 2023. The price for the M is £1,995 (roughly $2,400), or as a subscription product for £69 ($83) per month. Despite the fact that the product isn’t going to be available for more than half a year, there’s been a surprising amount of people pre-ordering a model. And we can expect to see plenty more models from the company if the M becomes a big seller. “Tesla didn’t launch with the Model Three,” said CTO Harry Wills, and like most companies, plans to start with a high-end offering before producing increasingly more affordable models. “Starting right, with a group of customers that adore the product, and then bringing it out to more accessible price points is what’s really important to us,” said Wills.

Of course, the economics of buying (or subscribing) to a scooter for upward of two grand is different compared to a £399 Xiaomi. It’s a serious, long term investment, and something that the pair are determined to ensure they can justify. “No-one trusts Bo yet,” said Morgan, “we’re a new company, and so we have to give [our customers] a great experience.” That means a generous warranty scheme, at least for the earliest adopters, and plenty of contact with the team to help iron out any kinks.

Daniel Cooper

And plenty of effort has been dedicated to making sure that M will run for years and years. Wills explained that “someone with very basic tools – an Allen key and the ability to plug and unplug – should be able to swap the battery if they so choose.” He added that nobody should expect to need to make such a drastic repair for many years, but that when the time comes, it’ll be easy for anyone to achieve. Wills said that the team has sweated the fine details to ensure that “all of the cables are in the right place and the connectors are in the right place.” Not to mention that a product that’s easier to maintain is also often easier to build, which should reduce costs on the manufacturing side.

Neither Morgan nor Wills are new to the world of scooter manufacture, and both worked for British e-mobility retailer and manufacturer Pure Electric. This is not their first scooter product, and they’re taking lessons they’ve already learned. Wills said that early buyers have signed up to the “ride quality, the safety and the overall aesthetics of the product.” In fact, Wills said that people have walked over to them and offered cash just to own the prototypes.

Daniel Cooper

I spent longer than I’d planned riding the two Bo M units up and down a private track in London. (Scooters which aren’t part of a licensed scooter-sharing trial remain illegal to ride on public roads in the UK, although that’s expected to change in 2023.) The first thing to say is that the benefits of Safesteer are obvious from the second you step onto the deck and push off. Unlike most regular scooters, which always feel like they are seconds from disaster, this felt solidly planted on the road. The grip and ability to turn at very low speeds is useful for weaving through narrow spaces (and sliding through some tightly-packed anti-traffic bollards).

One of my bad habits when riding most standard e-scooters is that I’ll bend my knees to try and lower my center of gravity. I’m sufficiently paranoid about taking a tumble that I never want to just stand up and ride the thing as you’re meant to. But it took me about half a minute before I realized that I didn’t just have to stand up on Bo M, but I could actually relax, stop gripping the handles so tightly and enjoy the ride, and the scenery. It’s an experience that is sufficiently car- like that I have to agree with the claim that it’s the sort of scooter only a group of “serious car guys” could build.

Morgan said that the other thing that they liked doing was building-in surprises for its would-be user base. The seamless body and flat, unmarked top exudes class, even in this early, 3D-printed-component state. Wills asked me to identify the power on / off switch, something that I proceeded to work my way through the scooter’s body, and down towards its deck, without much success. After a minute or so of fruitless hunting, Wills reached over and powered the unit down in a way I wouldn’t have spotted if I’d been there an hour (I’m sworn to secrecy as to where the switch is). If the company can offer the same level of surprise and delight to all of its users next year, then it’ll be off to a very strong start.

Is there much more space for innovation in desks, I hear you ask? After all, now that we’ve made them go up and down, there are no new worlds worth conquering. Not so, says Lumina, makers of its eponymous AI webcam that’s been described as the equal of a DSLR in some corners. Now, the company is turning its attention to building a smart desk with a programmable, 24-inch OLED screen nestled in its top, designed to offer you a place to put passive data in easy view.

The Lumina Desk, as it’s called, is a powered sit-stand desk, with a display that sits between your laptop (or monitor) and your keyboard and mouse. The idea, as you can see, is to offer you space to add in a view for your Google Calendar, stock view from Robinhood or a Twitter feed. All things that it’d be nice to always have in view, but you may not necessarily want to buy a second (or third) monitor for.

Lumina

Rather than a touchscreen, you’ll control data from a companion app on the desktop, and the whole desk is coated in a layer of toughened, anti-glare glass. The company is also promising a maximum brightness of 1,200 nits, and a 60Hz refresh rate, making it a pretty sweet display all told. There’s also plenty of power coursing through its frame, with two 20 x 20 Qi wireless charging pads on the tabletop, each one able to pump out 100W. Then there six hidden AC outlets with a built-in circuit breaker, and six USB-C ports, each one able to deliver up to 30W.

Sadly, you won’t be able to get your hands on one of these until 2023 at the earliest, and the company is today opening up unpaid reservations to gauge future interest. The price, too, has yet to be decided (although I’m told it’ll be as close to $1,000 as possible, at least that’s the aim right now) but if you’ve been looking for a more sci-fi way to passively collect information, this might be worth a peep.

We’ve known since early July that Samsung was preparing a gentle upgrade to its Galaxy Watch series this time around. After all, the Watch 4s were the first to get the new-and-improved WearOS after Google and Samsung decided to collaborate on its development. Consequently, the words of the day are evolution and refinement on the Watch 5, rather than anything too gee-whizz, with better sensors used to drive better data, and therefore give you a clearer picture of your overall health.

Both watches, for instance, get a reshaped back design with a larger surface area for better contact with your wrist. Samsung says the tweak will offer more accurate readings for your vital statistics, alongside a new infrared temperature sensor which can measure the ambient heat as well as what your skin’s throwing out. The BioActive sensor which drives many of the health features on the watch first made its debut on the Watch 4, but Samsung says the data it provides is far smarter now.

For instance, users can expect to see snapshots of their overall health, as well as personalized workouts that’ll help them get closer to their goals. Hell, you’ll even get recommendations on how much water to drink when you’re cooling down, based on how much sweat the watch thinks you’ve produced during your efforts. Similarly, a month-long guided sleep program is designed to help you get better rest (and will tailor your home accordingly if you’ve got the right gear).

The watch crystal has been swapped out for a tougher, sapphire crystal glass that Samsung says is significantly stronger than its predecessor. The batteries in both the 40mm (284mAh) and 44mm (410mAh) Watch 5 models are bigger, and both will get enough juice to track eight hours’ of sleep with just eight minutes on the charging pad. Both will run Wear OS 3.5, with One UI 4.5 sat on top, which has already been floating around for about a month or so at this point.

Samsung

Of course, Samsung also has a watch just for the more serious adventurers among us, the Watch 5 Pro. The big difference between the Pro and its vanilla siblings is the materials used in its construction, with a titanium body and a sport band. What the Pro has going for it otherwise is its bigger 45mm display and its longevity, since there’s a 590mAh battery nestled inside. You’ll also get some smarter navigation features for when you’re out and about and want to rely on your watch when your sense of direction might not cut it.

Now, let’s talk price. The Watch 5 will set you back $280 for the 40mm Bluetooth version, and $330 for the LTE model. The Watch 5 Pro, meanwhile, costs $450 for Bluetooth, and $500 if you want LTE thrown in for your trouble. If you’re the sort of person who has opinions about averages, tees and, uh, I’ve run out of Golf terms, you can also get a Galaxy Watch 5 Golf Edition. That ships in the same case sizes as the regular Watch 5s, but you’ll get a different strap, custom Golf-themed faces and an unlimited subscription to the Smart Caddie app thrown in.

Samsung is also throwing in a bunch of retail incentives for pre-order customers, including a Wireless Charger Duo if you lay down cash early for a Watch 5 Pro. It’s expected to hit store shelves on August 26th, and the company adds that you can also use Samsung’s bespoke studio to customize your phone and watch styles if you’re looking for something a bit more personal.

Follow all of the news from Samsung's Unpacked event right here!

Anglo-American EV startup Arrival is putting its groundbreaking bus and car projects on ice as it struggles to manage its cash reserves. The Financial Times reports that the company, which said it would lay off a third of its staff last month, would now focus on completing its delivery van. Arrival said that it had anything up to 20,000 orders with UPS for the vehicle, and is expecting to get the first models out of the door later this year. That will hopefully reduce the pressure on the company’s bottom line, and boost its share price, which has fallen 90 percent since it went public via a SPAC last year.

The company was unable to comment to Engadget about the FT’s report, as it is preparing to release its financial results this week.

Arrival actually started with its electric bus project, and has already built several models ready for real-world testing. Its car, designed to be sold to ride-share drivers, was at the prototype stage (I saw it first hand last December), and the company had recruited Tom Elvidge from Uber to run the program. The FT’s report says that both projects are in stasis for now, and are likely to be revived as soon as Arrival begins making money. The car project may, however, find itself squeezed by the looming recession and that so much VC money, which was dumped into transportation startups like Uber, has now dried up, leading to a wave of closures.

The biggest tragedy from all of this is that Arrival’s focus on revolutionizing public buses was a genuinely different approach from most EV makers. Buses are a fixture in pretty much every city, and while it’s always better for the environment to use one over a car, making them even cleaner was a great plan. That the public project has been iced in favor of the fleet of logistics vans is not surprising, but it’s certainly not a great sign for the future of public transport.

The theory and practice of marginal gains is to find and fix hundreds of small things that, in aggregate, add up to something vast. Asics believes that there are gains to be made in what runners wear when they’re at home as much as what they’re wearing on the track. That’s the pitch for the Actibreeze 3D, a pair of 3D-printed sandals with a lattice structure designed to improve cooling and breathability. The idea is to stop your extremities from getting too sweaty and tense after a run, so you’re that much more prepared for your next one. I’ve been wearing a pair for a couple of days now, and while they do keep your feet cool and dry, they’re not perfect.

Taking them out of the box, you’ll first notice how heavy they are, with each sandal – although they’re more like slippers – weighing 350 grams (12 oz) for my size 11s. They’re a lot bigger than your average pool slide, too, thanks to the overbuilt sole and lattice going over the top of your foot. Obviously, this is to help get air flowing under your feet to cool them down after a long run, and I experienced this after a fairly intensive gym session. It helped that we’re enduring a climate change-enhanced heatwave right now, to really ram home the lack of sweating. It’s a far nicer experience wearing these than what I’d normally use, which is a $15 pair of Havaianas.

Daniel Cooper

The 3D lattice is designed to provide the maximum amount of “step-in comfort” available, which means they’re pretty bouncy. Not in a I’m-walking-on-air way, but in that whenever you step, you can feel the sole compressing and bouncing back as you walk. I don’t know if the effect is more pronounced here than on other 3D-printed soles on the market, or if it’s magnified because you’re barefoot rather than wearing socks. Certainly, it takes a little mental calibration to compensate for the level of travel you’ll experience during each step. Maybe those folks who wear those novelty moon boots will find these no big deal, but if you’re coming from something flat, it is a noticeable change.

Here’s the issue – obviously 3D-printed stuff is made of springy plastic, but it’s still plastic, with its mostly hard, not-particularly-yielding structure. Wear these for an hour and the soles of your feet will look like you’ve been standing on a colander, the skin covered in a grid of little squares. Whatever benefits your feet are getting on the macro level, it requires you to tolerate the small annoyance of having your skin fed through a mesh. And, on a similar theme, because it’s a hard, waterproof plastic, it’s not the ideal surface to put your feet in close contact during a heatwave. That’s perhaps the one area that my $15 Havaianas have the edge, since there’s so little material coming into contact with the top of my feet. But if you're only wearing these for the two or three hours after you've had a running session, that shouldn't be too much of a problem. 

Daniel Cooper

Asics’ Actibreeze 3D are listed on the company’s website for $80, although they are currently not out for delivery. The company tells me that the stock will be available in selected markets once again this Autumn. 

Framework sold its eponymous laptop on the promise that end users should feel comfortable enough to fix almost any hardware problem themselves. Replacing a component shouldn’t be the reserve of dedicated service professionals if all you need is a T5 screwdriver and patience.

When the company released its new 12th-generation Intel Core mainboards, it couriered over a new board which could be inserted into last year’s model. And given that I don’t consider myself to be a very confident DIY-er, it made sense for me to put Framework’s promises to the test.

As you can see in the video below, laptops aren’t yet at the stage where you can pull components out as if they were Lego bricks. Although I think the industry is missing a trick by not making these components a lot simpler to assemble by standardizing the connections.

That said, one of the biggest hurdles was the ZIF connectors, which briefly made me wonder if I was really cut out for tech journalism. Those little lay-flat ribbons may be great for space but they’re a nightmare if you’ve got big hands and poor eyesight.

But, as I said when the new hardware was released, it’s empowering just how easy this stuff can be, more or less, if companies make even the smallest bit of effort. I was able to do this, talk to the camera (which probably slowed me down by quite a bit) and not blow myself up. And if I can do this, then surely you can too.

In 1917, Marcel Duchamp submitted a sculpture to the Society of Independent Artists under a false name. Fountain was a urinal, bought from a toilet supplier, with the signature R. Mutt on its side in black paint. Duchamp wanted to see if the society would abide by its promise to accept submissions without censorship or favor. (It did not.) But Duchamp was also looking to broaden the notion of what art is, saying a ready-made object in the right context would qualify. In 1962, Andy Warhol would twist convention with Campbell’s Soup Cans, 32 paintings of soup cans, each one a different flavor. Then, as before, the debate raged about if something mechanically produced – a urinal, or a soup can (albeit hand-painted by Warhol) – counted as art, and what that meant.

Now, the debate has been turned upon its head, as machines can mass-produce unique pieces of art on their own. Generative Artificial Intelligences (GAIs) are systems which create pieces of work that can equal the old masters in technique, if not in intent. But there is a problem, since these systems are trained on existing material, often using content pulled from the internet, from us. Is it right, then, that the AIs of the future are able to produce something magical on the backs of our labor, potentially without our consent or compensation?

The new frontier

The most famous GAI right now is DALL-E 2, Open AI’s system for creating “realistic images and art from a description in natural language.” A user could enter the phrase “teddy bears shopping for groceries in the style of Ukiyo-e,” and the model will produce pictures in that style. Similarly, ask for the bears to be shopping in Ancient Egypt and the images will look more like dioramas from a museum depicting life under the Pharaohs. To the untrained eye, some of these pictures look like they were drawn in 17th-century Japan, or shot at a museum in the 1980s. And these results are coming despite the technology still being at a relatively early stage.

Open AI recently announced that DALL-E 2 would be made available to up to one million users as part of a large-scale beta test. Each user will be able to make 50 generations for free during their first month of use, and then 15 for every subsequent month. (A generation is either the production of four images from a single prompt, or the creation of three more if you choose to edit or vary something that’s already been produced.) Additional 115-credit packages can be bought for $15, and the company says more detailed pricing is likely to come as the product evolves. Crucially, users are entitled to commercialize the images produced with DALL-E, letting them print, sell or otherwise license the pictures borne from their prompts.

Open AI

These systems did not, however, develop an eye for a good picture in a vacuum, and each GAI has to be trained. Artificial Intelligence is, after all, a fancy term for what is essentially a way of teaching software how to recognize patterns. “You allow an algorithm to develop that can be improved through experience,” said Ben Hagag, head of research at Darrow, an AI startup looking to improve access to justice. “And by experience I mean examining and finding patterns in data.” “We say to the [system] ‘take a look at this dataset and find patterns,” which then go on to form a coherent view of the data at hand. “The model learns as a baby learns,” he said, so if a baby looked at a 1,000 pictures of a landscape, it would soon understand that the sky – normally oriented across the top of the image – would be blue while land is green.

Hagag cited how Google built its language model by training a system on several gigabytes of text, from the dictionary to examples of the written word. “The model understood the patterns, how the language is built, the syntax and even the hidden structure that even linguists find hard to define,” Hagag said. Now that model is sophisticated enough that “once you give it a few words, it can predict the next few words you’re going to write.” In 2018, Google’s Ajit Varma told The Wall Street Journal that its smart reply feature had been trained on “billions of Gmail messages,” adding that initial tests saw options like ‘I Love You’ and ‘Sent from my iPhone’ offered up since they were so commonly seen in communications.

Developers who do not have the benefit of access to a data set as vast as Google’s need to find data via other means. “Every researcher developing a language model first downloads Wikipedia then adds more,” Hagag said. He added that they are likely to pull down any, and every, piece of available data that they can find. The sassy tweet you sent a few years ago, or that sincere Facebook post, may have been used to train someone’s language model, somewhere. Even Open AI uses social media posts with WebText, a dataset which pulls text from outbound Reddit links which received at least three karma, albeit with Wikipedia references removed.

Guan Wang, CTO of Huski, says that the pulling down of data is “very common.” “Open internet data is the go-to for the majority of AI model training nowadays,” he said. And that it’s the policy of most researchers to get as much data as they can. “When we look for speech data, we will get whatever speech we can get,” he added. This policy of more data-is-more is known to produce less than ideal results, and Ben Hagag cited Riley Newman, former head of data science at Airbnb, who said “better data beats more data,” but Hagag notes that often, “it’s easier to get more data than it is to clean it.”

Craiyon / Daniel Cooper

DALL-E may now be available to a million users, but it’s likely that people’s first experience of a GAI is with its less-fancy sibling. Craiyon, formerly DALL-E Mini, is the brainchild of French developer Boris Dayma, who started work on his model after reading Open AI’s original DALL-E paper. Not long after, Google and the AI development community HuggingFace ran a hackathon for people to build quick-and-dirty machine learning models. “I suggested, ‘Hey, let’s replicate DALL-E. I have no clue how to do that, but let’s do it,” said Dayma. The team would go on to win the competition, albeit with a rudimentary, rough-around-the-edges version of the system. “The image [it produced] was clear. It wasn’t great, but it wasn’t horrible,” he added. But unlike the full-fat DALLl-E, Dayma’s team was focused on slimming the model down so that it could work on comparatively low-powered hardware.

Dayma’s original model was fairly open about which image sets it would pull from, often with problematic consequences. “In early models, still in some models, you ask for a picture – for example mountains under the snow,” he said, “and then on top of it, the Shutterstock or Alamy watermark.” It’s something many AI researchers have found, with GAIs being trained on those image libraries public-facing image catalogs, which are covered in anti-piracy watermarks.

Dayma said that the model had erroneously learned that high-quality landscape images typically had a watermark from one of those public photo libraries, and removed them from his model. He added that some early results also output not-safe-for-work responses, forcing him to make further refinements to his initial training set. Dayma added that he had to do a lot of the sorting through the data himself, and said that “a lot of the images on the internet are bad.”

But it’s not just Dayma who has noticed the regular appearance of a Shutterstock watermark, or something a lot like it, popping up in AI-generated art. Which begs the question, are people just ripping off Shutterstock’s public-facing library to train their AI? It appears that one of the causes is Google, which has indexed a whole host of watermarked Shutterstock images as part of its Conceptual Captions framework. Delve into the data, and you’ll see a list of image URLs which can be used to train your own AI model, thousands of which are from Shutterstock. Shutterstock declined to comment on the practice for this article.

A Google spokesperson said that they don’t “believe this is an issue for the datasets we’re involved with.” They also quoted from this Creative Commons report, saying that “the use of works to train AI should be considered non-infringing by default, assuming that access to the copyright works was lawful at the point of input.” That is despite the fact that Shutterstock itself expressly forbids visitors to its site from using “any data mining, robots or similar data and/or image gathering and extraction methods in connection with the site or Shutterstock content.”

Alex Cardinelli, CEO at AI startup Article Forge, says that he sees no issue with models being trained on copyrighted texts, “so long as the material itself was lawfully acquired and the model does not plagiarize the material.” He compared the situation to a student reading the work of an established author, who may “learn the author’s styles or patterns, and later find applicable places to reuse those concepts.” He added that so long as a model isn’t “copying and pasting from their training data,” then it simply repeats a pattern that has appeared since the written word began.

Dayma says that, at present, hundreds of thousands, if not millions of people are playing with his system on a daily basis. That all incurs a cost, both for hosting and processing, which he couldn’t sustain from his own pocket for very long, especially since it remains a “hobby.” Consequently, the site runs ads at the top and bottom of its page, between which you’ll get a grid of nine surreal images. “For people who use the site commercially, we could always charge for it,” he suggested. But he admitted his knowledge of US copyright law wasn’t detailed enough to be able to discuss the impact of his own model, or others in the space. This is the situation that Open AI also perhaps finds itself dealing with given that it is now allowing users to sell pictures created by DALL-E.

The law of art

The legal situation is not a particularly clear one, especially not in the US, where there have been few cases covering Text and Data Mining, or TDM. This is the technical term for the training of an AI by plowing through a vast trove of source material looking for patterns. In the US, TDM is broadly covered by Fair Use, which permits various forms of copying and scanning for the purposes of allowing access. This isn’t, however, a settled subject, but there is one case that people believe sets enough of a precedent to enable the practice.

Authors Guild v. Google (2015) was brought by a body representing authors, which accused Google of digitizing printed works that were still held under copyright. The initial purpose of the work was, in partnership with several libraries, to catalog and database the texts to make research easier. Authors, however, were concerned that Google was violating copyright, and even if it wasn’t making the text of a still-copyrighted work available publicly, it was prohibited from scanning and storing it in the first place. Eventually, the Second Circuit ruled in favor of Google, saying that digitizing copyright-protected work did not constitute copyright infringement.

Rahul Telang is Professor of Information Systems at Carnegie Mellon University, and an expert in digitization and copyright. He says that the issue is “multi-dimensional,” and that the Google Books case offers a “sort of precedent” but not a solid one. “I wish I could tell you there was a clear answer,” he said, “but it’s a complicated issue,” especially around works that may or may not be transformative. And until there is a solid case, it’s likely that courts will apply the usual tests for copyright infringement, around if a work supplants the need for the original, and if it causes economic harm to the original rights holder. Telang believes that countries will look to loosen restrictions on TDM wherever possible in order to boost domestic AI research.

The US Copyright Office says that it will register an “original work of authorship, provided that the work was created by a human being.” This is due to the old precedent that the only thing worth copyrighting is “the fruits of intellectual labor,” produced by the “creative powers of the mind.” In 1991, this principle was affirmed by a case of purloined listings from one phone book company by another. The Supreme Court held that while effort may have gone into the compilation of a phone book, the information contained therein was not an original work, created by a human being, and so therefore couldn’t be copyrighted. It will be interesting to see if there are any challenges made to users trying to license or sell a DALL-E work for this very reason.

Rob Holmes, a private investigator who works on copyright and trademark infringement with many major tech companies and fashion brands, believes that there is a reticence across the industry to pursue a landmark case that would settle the issue around TDM and copyright. “Legal departments get very little money,” he said. “All these different brands, and everyone’s waiting for the other brand, or IP owner, to begin the lawsuit. And when they do, it’s because some senior VP or somebody at the top decided to spend the money, and once that happens, there’s a good year of planning the litigation.” That often gives smaller companies plenty of time to either get their house in order, get big enough to be worth a lawsuit or go out of business.

“Setting a precedent as a sole company costs a lot of money,” Holmes said, but brands will move fast if there’s an immediate risk to profitability. Designer brand Hermés, for instance, is suing an artist named Mason Rothschild, who is producing MetaBirkins NFTs. These are styled images on a design reminiscent of Hermés’ famous Birkin handbag, something the French fashion house says is nothing more than an old-fashioned rip-off. This, too, is likely to have ramifications for the industry as it wrestles with philosophical questions of what work is sufficiently transformational as to prevent an accusation of piracy.

Artists are also able to upload their own work to DALL-E and then generate recreations in their own style. I spoke to one artist, who asked not to be named or otherwise described for fear of being identified and suffering reprisals. They showed me examples of their work alongside recreations made by DALLl-E, which while crude, were still close enough to look like the real thing. They said that, on this evidence alone, their livelihood as a working artist is at risk, and that the creative industries writ large are “doomed.”

Article Forge CEO Alex Cardinelli says that this situation, again, has historical precedent with the industrial revolution. He says that, unlike then, society has a collective responsibility to “make sure that anyone who is displaced is adequately supported.” And that anyone in the AI space should be backing a “robust safety net,” including “universal basic income and free access to education,” which he says is the “bare minimum” a society in the midst of such a revolution should offer.

Trained on your data

AIs are already in use. Microsoft, for instance, partnered with OpenAI to harness GPT-3 as a way to build code. In 2021, the company announced that it would integrate the system into its low-code app-development platform to help people build apps and tools for Microsoft products. Duolingo uses the system to improve people’s French grammar, while apps like Flowrite employs it to help make writing blog posts and emails easier and faster. Midjourney, a DALL-E 2-esque GAI for art, which has recently opened up its beta, is capable of producing stunning illustrated art – with customers charged between $10-50 a month if they wish to produce more images or use those pictures commercially.

For now, that’s something Craiyon doesn't necessarily need to worry about, since the resolution is presently so low. “People ask me ‘why is the model bad on faces’, not realizing that the model is equally good – or bad – at everything,” Dayma said. “It’s just that, you know, when you draw a tree, if the leaves are messed up you don’t care, but when the faces or eyes are, we put more attention on it.” This will, however, take time both to improve the model, and to improve the accessibility of computing power capable of producing the work. Dayma believes that despite any notion of low quality, any GAI will need to be respectful of “the applicable laws,” and that it shouldn’t be used for “harmful purposes.”

And artificial intelligence isn’t simply a toy, or an interesting research project, but something that has already caused plenty of harm. Take Clearview AI, a company that scraped several billion images, including from social media platforms, to build what it claims is a comprehensive image recognition database. According to The New York Times, this technology was used by billionaire John Catsimatidis to identify his daughter’s boyfriend. BuzzFeed News reported that Clearview has offered access not just to law enforcement – its supposed corporate goal – but to a number of figures associated with the far right. The system has also proved less than reliable, with The Times reporting that it has led to a number of wrongful arrests.

Naturally, the ability to synthesize any image without the need for a lot of photoshopping should raise alarm. Deepfakes, a system that uses AI to replace someone’s face in a video has already been used to produce adult content featuring celebrities. As quickly as companies making AIs can put in guardrails to prevent adult-content prompts, it’s likely that loopholes will be found. And as open-source research and development becomes more prevalent, it’s likely that other platforms will be created with less scrupulous aims. Not to mention the risk of this technology being used for political ends, given the ease of creating fake imagery that could be used for propaganda purposes.

Of course, Duchamp and Warhol may have stretched the definitions of what art can be, but they did not destroy art in and of itself. It would be a mistake to suggest that automating image generation will inevitably lead to the collapse of civilization. But it’s worth being cautious about the effects on artists, who may find themselves without a living if it’s easier to commission a GAI to produce something for you. Not to mention the implication for what, and how, these systems are creating material for sale on the backs of our data. Perhaps it is time that we examined if it’s necessary to implement a way of protecting our material – something equivalent to Do Not Track – to prevent it being chewed up and crunched through the AI sausage machine.

Spotify’s second-quarter financial release shows the streaming giant hasn’t yet felt the dread hand of the looming global recession. Unlike Netflix, which had to report a fall in its overall customer base, Spotify has seen both free and paying accounts grow. It now has 433 million users, up from the 422 million reported at the end of the first quarter. 188 million of those are paying for Premium, a leap of six million from three months ago, while a further four million are signed up on an ad-supported basis.

Despite industry-wide fears that household budgets would cut entertainment costs to help free up much-needed cash, Spotify has dodged cost-cutting so far. The company said that while it was keeping an eye on the “uncertain” environment, it was “pleased with the resilience of [its] business.” That said, the company did spend big to help grow its user figures, with marketing campaigns designed to coax back users who let their subscriptions lapse, or who wanted to expand to a family plan.

That marketing spend helped blow a hole in the company’s finances, with Spotify posting a quarterly loss of €194 million ($197 million). The company is banking on sharp increases in revenue both for subscriptions and advertising to help balance those losses out. In addition, its plan to pivot toward cheaper forms of audio content, like podcasts and audiobooks, should see the volume of cash it pays to record labels fall to a more tolerable (for Spotify) level — even if recording artists continue to demonstrate that they’re being starved of an income by the piddling royalties paid out on a per-stream basis.

Twitter has today announced its results for the second-quarter of 2022, saying that it has seen a sharp rise in the number of regular users. In the last three months, Monetizable Daily Active Users (mDAU) climbed from 39.6 million to 41.5 million, while global reach leapt from 189.4 million in April to 196.3 million today. Unfortunately, those increasing user figures did not see a boost in the company’s bottom line, and revenue was $1.18 billion, which is slightly down both year-on-year and quarter-on-quarter.

Worse still for a company bringing in that much revenue is that costs and expenses for the period equalled $1.52 billion, with extra pain coming from both the costs of dealing with Elon Musk’s purchase of Twitter and paying severance for all of the workers it’s been laying off as part of its cost-cutting drive. All in all, the company posted a net loss of $270 million, much of which it attributes to both the looming recession and the uncertainty around the proposed takeover.

Back in April, as part of its first quarter financial release, Twitter revealed that it had historically miscounted its user figures. Between 2019 and 2021, the company had counted users with multiple accounts as multiple people, adding up to two million users to the figures. This, while not a catastrophic admission, did serve to highlight that Twitter’s slow growth was even slower than people believed. At the time, the company also said that it had earned $1.20 billion in revenue, $1.11 billion of which was produced through advertising, while the average monetizable daily user figures hit 39.6 million in the US and 189.4 million in the rest of the world.

While this was going on, Twitter had also been targeted as an acquisition vehicle for Elon Musk, and the deal has dominated much of the news cycle ever since. The Tesla and SpaceX CEO pledged to buy the company at a very high valuation, and signed a binding agreement that opted to waive much of the due diligence often necessary in deals like this. Not long after, however, Musk decided — either on his own, or influenced by Tesla’s dwindling stock price — to try and pull out of the deal, claiming that Twitter had misrepresented how many automated accounts were on the platform.

Unfortunately for Musk, contract law is often funny about letting people walk away from deals they signed promising to waive the necessary due diligence. Twitter has since sued the figure in order to either force him to buy, or to pay a significant sum to make the whole thing go away. The Delaware Court of Chancery rejected Musk’s request to hold a trial in 2023, and accepted Twitter’s plea to expedite the matter. Consequently, the pair will square off for a five-day courtroom showdown in October.

Twitter has said, once again, that it believes Musk’s “purported termination is invalid and wrongful,” and that the proposed merger deal “remains in effect.”