After initially disabling the capability, OpenAI today announced that customers with access to DALL-E 2 can upload people’s faces to edit them using the AI-powered image-generating system. Previously, OpenAI only allowed users to work with and share photorealistic faces and banned the uploading of any photo that might depict a real person, including photos of prominent celebrities and public figures.
OpenAI claims that improvements to its safety system made the face-editing feature possible by “minimizing the potential of harm” from deepfakes as well as attempts to create sexual, political and violent content. In an email to customers, the company wrote:
Many of you have told us that you miss using DALL-E to dream up outfits and hairstyles on yourselves and edit the backgrounds of family photos. A reconstructive surgeon told us that he’d been using DALL-E to help his patients visualize results. And filmmakers have told us that they want to be able to edit images of scenes with people to help speed up their creative processes … [We] built new detection and response techniques to stop misuse.
The change in policy isn’t opening the floodgates necessarily. OpenAI’s terms of service will continue to prohibit uploading pictures of people without their consent or images that users don’t have the rights to — although it’s not clear how consistent the company’s historically been about enforcing those policies.
In any case, it’ll be a true test of OpenAI’s filtering technology, which some customers in the past have complained about being overzealous and somewhat inaccurate. Deepfakes come in many flavors, from fake vacation photos to presidents of war-torn countries. Accounting for every emerging form of abuse will be a never-ending battle, in some cases with very high stakes.
No doubt, OpenAI — which has the backing of Microsoft and notable VC firms including Khosla Ventures — is eager to avoid the controversy associated with Stability AI’s Stable Diffusion, an image-generating system that’s available in an open source format without any restrictions. As TechCrunch recently wrote about, it didn’t take long before Stable Diffusion — which can also edit face images — was being used by some to create pornographic, nonconsensual deepfakes of celebrities like Emma Watson.
So far, OpenAI has positioned itself as a brand-friendly, buttoned-up alternative to the no-holds-barred Stability AI. And with the constraints around the new face editing feature for DALL-E 2, the company is maintaining the status quo.
DALL-E 2 remains in invite-only beta. In late August, OpenAI announced that over a million people are using the service.
OpenAI begins allowing users to edit faces with DALL-E 2 by Kyle Wiggers originally published on TechCrunch
OpenAI begins allowing users to edit faces with DALL-E 2
Архив рубрики: Artificial Intelligence
AI is getting better at generating porn. We might not be prepared for the consequences.
A red-headed woman stands on the moon, her face obscured. Her naked body looks like it belongs on a poster you’d find on a hormonal teenager’s bedroom wall — that is, until you reach her torso, where three arms spit out of her shoulders.
AI-powered systems like Stable Diffusion, which translate text prompts into pictures, have been used by brands and artists to create concept images, award-winning (albeit controversial) prints and full-blown marketing campaigns.
But some users, intent on exploring the systems’ murkier side, have been testing them for a different sort of use case: porn.
AI porn is about as unsettling and imperfect as you’d expect (that red-head on the moon was likely not generated by someone with an extra arm fetish). But as the tech continues to improve, it will evoke challenging questions for AI ethicists and sex workers alike.
Pornography created using the latest image-generating systems first arrived on the scene via the discussion boards 4chan and Reddit earlier this month, after a member of 4chan leaked the open source Stable Diffusion system ahead of its official release. Then, last week, what appears to be one of the first websites dedicated to high-fidelity AI porn generation launched.
Called Porn Pen, the website allows users to customize the appearance of nude AI-generated models — all of which are women — using toggleable tags like “babe,” “lingerie model,” “chubby,” ethnicities (e.g. “Russian” and “Latina”) and backdrops (e.g. “bedroom,” “shower” and wildcards like “moon”). Buttons capture models from the front, back or side, and change the appearance of the generated photo (e.g. “film photo,” “mirror selfie”). There must be a bug on the mirror selfies, though, because in the feed of user-generated images, some mirrors don’t actually reflect a person — but of course, these models are not people at all. Porn Pen functions like “This Person Does Not Exist,” only it’s NSFW.
On Y Combinator’s Hacker News forum, a user purporting to be the creator describes Porn Pen as an “experiment” using cutting-edge text-to-image models. “I explicitly removed the ability to specify custom text to avoid harmful imagery from being generated,” they wrote. “New tags will be added once the prompt-engineering algorithm is fine-tuned further.” The creator did not respond to TechCrunch’s request for comment.
But Porn Pen raises a host of ethical questions, like biases in image-generating systems and the sources of the data from which they arose. Beyond the technical implications, one wonders whether new tech to create customized porn — assuming it catches on — could hurt adult content creators who make a living doing the same.
“I think it’s somewhat inevitable that this would come to exist when [OpenAI’s] DALL-E did,” Os Keyes, a PhD candidate at Seattle University, told TechCrunch via email. “But it’s still depressing how both the options and defaults replicate a very heteronormative and male gaze.”
Ashley, a sex worker and peer organizer who works on cases involving content moderation, thinks that the content generated by Porn Pen isn’t a threat to sex workers in its current state.
“There is endless media out there,” said Ashley, who did not want her last name to be published for fear of being harassed for their job. “But people differentiate themselves not by just making the best media, but also by being an accessible, interesting person. It’s going to be a long time before AI can replace that.”
On existing monetizable porn sites like OnlyFans and ManyVids, adult creators must verify their age and identity so that the company knows they are consenting adults. AI-generated porn models can’t do this, of course, because they aren’t real.
Ashley worries, though, that if porn sites crack down on AI porn, it might lead to harsher restrictions for sex workers, who are already facing increased regulation from legislation like SESTA/FOSTA. Congress introduced the Safe Sex Workers Study Act in 2019 to examine the affects of this legislation, which makes online sex work more difficult. This study found that “community organizations [had] reported increased homelessness of sex workers” after losing the “economic stability provided by access to online platforms.”
“SESTA was sold as fighting child sex trafficking, but it created a new criminal law about prostitution that had nothing about age,” Ashley said.
Currently, few laws around the world pertain to deepfaked porn. In the U.S., only Virginia and California have regulations restricting certain uses of faked and deepfaked pornographic media.
Systems such as Stable Diffusion “learn” to generate images from text by example. Fed billions of pictures labeled with annotations that indicate their content — for example, a picture of a dog labeled “Dachshund, wide-angle lens” — the systems learn that specific words and phrases refer to specific art styles, aesthetics, locations and so on.
This works relatively well in practice. A prompt like “a bird painting in the style of Van Gogh” will predictably yield a Van Gogh-esque image depicting a bird. But it gets trickier when the prompts are vaguer, refer to stereotypes or deal with subject matter with which the systems aren’t familiar.
For example, Porn Pen sometimes generates images without a person at all — presumably a failure of the system to understand the prompt. Other times, as alluded to earlier, it shows physically improbable models, typically with extra limbs, nipples in unusual places and contorted flesh.
“By definition [these systems are] going to represent those whose bodies are accepted and valued in mainstream society,” Keyes said, noting that Porn Pen only has categories for cisnormative people. “It’s not surprising to me that you’d end up with a disproportionately high number of women, for example.”
While Stable Diffusion, one of the systems likely underpinning Porn Pen, has relatively few “NSFW” images in its training dataset, early experiments from Redditors and 4chan users show that it’s quite competent at generating pornographic deepfakes of celebrities (Porn Pen — perhaps not coincidentally — has a “celebrity” option). And because it’s open source, there’d be nothing to prevent Porn Pen’s creator from fine-tuning the system on additional nude images.
“It’s definitely not great to generate [porn] of an existing person,” Ashley said. “It can be used to harass them.”
Deepfake porn is often created to threaten and harass people. These images are almost always developed without the subject’s consent out of malicious intent. In 2019, the research company Sensity AI found that 96% of deepfake videos online were non-consensual porn.
Mike Cook, an AI researcher who’s a part of the Knives and Paintbrushes collective, says that there’s a possibility the dataset includes people who’ve not consented to their image being used for training in this way, including sex workers.
“Many of [the people in the nudes in the training data] may derive their income from producing pornography or pornography-adjacent content,” Cook said. “Just like fine artists, musicians or journalists, the works these people have produced are being used to create systems that also undercut their ability to earn a living in the future.”
In theory, a porn actor could use copyright protections, defamation and potentially even human rights laws to fight the creator of a deepfaked image. But as a piece in MIT Technology Review notes, gathering evidence in support of the legal argument can prove to be a massive challenge.
When more primitive AI tools popularized deepfaked porn several years ago, a Wired investigation found that nonconsensual deepfake videos were racking up millions of views on mainstream porn sites like Pornhub. Other deepfaked works found a home on sites akin to Porn Pen — according to Sensity data, the top four deepfake porn websites received more than 134 million views in 2018.
“AI image synthesis is now a widespread and accessible technology, and I don’t think anyone is really prepared for the implications of this ubiquity,” Cook continued. “In my opinion, we have rushed very, very far into the unknown in the last few years with little regard for the impact of this technology.”
To Cook’s point, one of the most popular sites for AI-generated porn expanded late last year through partner agreements, referrals and an API, allowing the service — which hosts hundreds of nonconsensual deepfakes — to survive bans on its payments infrastructure. And in 2020, researchers discovered a Telegram bot that generated abusive deepfake images of more than 100,000 women, including underage girls.
“I think we’ll see a lot more people testing the limits of both the technology and society’s boundaries in the coming decade,” Cook said. “We must accept some responsibility for this and work to educate people about the ramifications of what they are doing.”
AI is getting better at generating porn. We might not be prepared for the consequences.
Datch secures $10M to build voice assistants to factory floors
Datch, a company that develops AI-powered voice assistants for industrial customers, today announced that it raised $10 million in a Series A round led by Blackhorn Ventures. The proceeds will be used to expand operations, CEO Mark Fosdike said, as well as develop new software support, tools and capabilities.
Datch started when Fosdike, who has a background in aerospace engineering, met two former Siemens engineers — Aric Thorn and Ben Purcell. They came to the collective realization that voice products built for business customers have to overcome business-specific challenges, like understanding jargon, acronyms and syntax unique to particular customers.
“The way we extract information from systems changes every year, but the way we input information — especially in the industrial world — hasn’t changed since the invention of the keyboard and database,” Fosdike said. “The industrial world had been left in the dark for years, and we knew that developing a technology with voice-visual AI would help light the way for these factories.”
The voice assistants that Datch builds leverage AI to collect and structure data from users in a factory or in the field, parsing commands like “Report an issue for the Line 1 Spot Welder. I estimate it will take half a day to fix.” They run on a smartphone and link to existing systems to write and read records, including records from enterprise resource and asset management platforms.
Datch’s assistants provide a timeline of events and can capture data without an internet connection; they auto-sync once back online. Using them, workers can fill out company forms, create and update work orders, assign tasks and search through company records all via voice.
Fosdike didn’t go into detail about how Datch treats the voice data, save that it encrypts data both in-transit and at rest and performs daily backups.
“We have to employ a lot of tight, automated feedback loops to train the voice and [language] data, and so everyone’s interaction with Datch is slightly different, depending on the company and team they work within,” Fosdike explained. “Customers are exploring different use cases such as using the [language] data in predictive maintenance, automated classification of cause codes, and using the voice data to predict worker fatigue before it becomes a critical safety risk.”
That last bit about predicting worker fatigue is a little suspect. The idea that conditions like tiredness can be detected in a person’s voice isn’t a new one, but some researchers believe it’s unlikely AI can flag them with 100% accuracy. After all, people express tiredness in different ways, depending not only on the workplace environment but on their sex and cultural, ethnic and demographic backgrounds.
The tiredness-detecting scenario aside, Fosdike asserts that Datch’s technology is helping industrial clients get ahead of turbulence in the economy by “vastly improving” the efficiency of their operations. Frontline staff typically have to work with reporting tools that aren’t intuitive, he notes, and in many cases, voice makes for a less cumbersome, faster alternative form of input.
“We help frontline workers with productivity and solve the pain point of time wasted on their reports by decreasing the process time,” Fosdike said. “Industrial companies are fast realizing that to keep up with demand or position themselves to withstand a global pandemic, they need to find a way to scale with more than just peoplepower. Our AI offers these companies an efficient solution in a fraction of the time and with less overhead needed.”
Datch competes with Rain, Aiqudo and Onvego, all of which are developing voice technologies for industrial customers. Deloitte’s Maxwell, Genba and Athena are rivals in Fosdike’s eyes, as as well. But business remains steady — Datch counts ConEd, Singapore Airlines, ABB Robotics and the New York Power Authority among its clients.
“We raised this latest round earlier than expected due to the influx of demand from the market. The timing is right to capitalize on both the post-COVID boom in digital transformation as well as corporate investments driven by the infrastructure bill,” Fosdike said, referring to the $1 trillion package U.S. lawmakers passed last November. “Currently we have a team of 20, and plan to use the funds to grow to 55 to 60 people, scaling to roughly 40 by the end of the year.”
To date, Datch has raised $15 million in venture capital.
Datch secures $10M to build voice assistants to factory floors
PayTalk promises to handle all sorts of payments with voice, but the app has a long way to go
Neji Tawo, the founder of boutique software development company Wiscount Corporation, says he was inspired by his dad to become an engineer. When Tawo was a kid, his dad tasked him with coming up with a formula to calculate the gas in the fuel tanks at his family’s station. Tawo then created an app for gas stations to help prevent gas siphoning.
The seed of the idea for Tawo’s latest venture came from a different source: a TV ad for a charity. Frustrated by his experience filling out donation forms, Tawo sought an alternative, faster way to complete such transactions. He settled on voice.
Tawo’s PayTalk, which is one of the first products in Amazon’s Black Founders Build with Alexa Program, uses conversational AI to carry out transactions via smart devices. Using the PayTalk app, users can do things like find a ride, order a meal, pay bills, purchase tickets and even apply for a loan, Tawo says.
“We see the opportunity in a generation that’s already using voice services for day-to-day tasks like checking the weather, playing music, calling friends and more,” Tawo said. “At PayTalk, we feel voice services should function like a person — being capable of doing several things from hailing you a ride to taking your delivery order to paying your phone bills.”
PayTalk is powered by out-of-the-box voice recognition models on the frontend and various API connectors behind the scenes, Tawo explains. In addition to Alexa, the app integrates with Siri and Google Assistant, letting users add voice shortcuts like “Hey Siri, make a reservation on PayTalk.”
“Myself and my team have bootstrapped this all along the way, as many VCs we approached early on were skeptical about voice being the device form factor of the future. The industry is in its nascent stages and many still view it with skepticism,” Tawo said. “With the COVID-19 pandemic and subsequent shift to doing more remotely across different types of transactions (i.e. ordering food from home, shopping online, etc.), we … saw that there was increased interest in the use of voice services. This in turn boosted demand for our product and we believe that we are positioned to continue to expand our offerings and make voice services more useful as a result.”
Tawo’s pitch for PayTalk reminded me much of Viv, the startup launched by Siri co-creator Adam Cheyer (later acquired by Samsung) that proposed voice as the connective tissue between disparate apps and services. It’s a promising idea — tantalizing, even. But where PayTalk is concerned, the execution isn’t quite there yet.
The PayTalk app is only available for iOS and Android at the moment, and in my experience with it, it’s a little rough around the edges. A chatbot-like flow allows you to type commands — a nice fallback for situations where voice doesn’t make sense (or isn’t appropriate) — but doesn’t transition to activities particularly gracefully. When I used it to look for a cab by typing the suggested “book a ride” command, PayTalk asked for a pickup and dropoff location before throwing me into an Apple Maps screen without any of the information I’d just entered.
The reservation and booking functionality seems broken as well. PayTalk walked me through the steps of finding a restaurant, asking which time I’d like to reserve, the size of my party and so on. But the app let me “confirm” a table for 2 a.m. at SS106 Aperitivo Bar — an Italian restaurant in Alberta — on a day the restaurant closes at 10 p.m.
Image Credits: PayTalk
Other “categories” of commands in PayTalk are very limited in what they can accomplish — or simply nonfunctional. I can only order groceries from two services in my area (Downtown Brooklyn) at present — MNO African Market and Simi African Foods Market. Requesting a loan prompts an email with a link to Glance Capital, a personal loan provider for gig workers, that throws a 404 error when clicked. A command to book “luxury services” like a yacht or “sea plane” (yes, really) fails to reach anything resembling a confirmation screen, while the “pay for parking” command confusingly asks for a zone number.
To fund purchases through PayTalk (e.g. parking), there’s an in-app wallet. I couldn’t figure out a way to transfer money to it, though. The app purports to accept payment cards, but tapping on the “Use Card” button triggers a loading animation that quickly times out.
I could go on. But suffice it to say that PayTalk is in the very earliest stages of development. I began to think the app had been released prematurely, but PayTalk’s official Twitter account has been advertising it for at least the past few months.
Perhaps PayTalk will eventually grow into the shoes of the pitch Tawo gave me, so to speak — Wiscount is kicking off a four-month tenure at the Black Founders Build with Alexa Program. In the meantime, it must be pointed out that Alexa, Google Assistant and Siri are already capable of handling much of what PayTalk promises to one day accomplish.
The battle for voice recognition inside vehicles is heating up
“With the potential $100,000 investment [from the Black Founders Build with Alexa Program], we will seek to raise a seed round to expand our product offerings to include features that would allow customers to seamlessly carry out e-commerce and financial transactions on voice service-powered devices,” Tawo said. “PayTalk is mainly a business-to-consumer platform. However, as we continue to innovate and integrate voice-activated options … we see the potential to support enterprise use cases by replacing and automating the lengthy form filling processes that are common for many industries like healthcare.”
Hopefully, the app’s basic capabilities get attention before anything else.
PayTalk promises to handle all sorts of payments with voice, but the app has a long way to go
How Niantic evolved Pokémon GO for the year no one could go anywhere
Pokémon GO was created to encourage players to explore the world while coordinating impromptu large group gatherings — activities we’ve all been encouraged to avoid since the pandemic began.
And yet, analysts estimate that 2020 was Pokémon GO’s highest-earning year yet.
By twisting some knobs and tweaking variables, Pokémon GO became much easier to play without leaving the house.
Niantic’s approach to 2020 was full of carefully considered changes, and I’ve highlighted many of their key decisions below.
Consider this something of an addendum to the Niantic EC-1 I wrote last year, where I outlined things like the company’s beginnings as a side project within Google, how Pokémon Go began as an April Fools’ joke and the company’s aim to build the platform that powers the AR headsets of the future.
Hit the brakes
On a press call outlining an update Niantic shipped in November, the company put it on no uncertain terms: the roadmap they’d followed over the last ten-or-so months was not the one they started the year with. Their original roadmap included a handful of new features that have yet to see the light of day. They declined to say what those features were of course (presumably because they still hope to launch them once the world is less broken) — but they just didn’t make sense to release right now.
Instead, as any potential end date for the pandemic slipped further into the horizon, the team refocused in Q1 2020 on figuring out ways to adapt what already worked and adjust existing gameplay to let players do more while going out less.
Turning the dials
As its name indicates, GO was never meant to be played while sitting at home. John Hanke’s initial vision for Niantic was focused around finding ways to get people outside and playing together; from its very first prototype, Niantic had players running around a city to take over its virtual equivalent block by block. They’d spent nearly a decade building up a database of real-world locations that would act as in-game points meant to encourage exploration and wandering. Years of development effort went into turning Pokémon GO into more and more of a social game, requiring teamwork and sometimes even flash mob-like meetups for its biggest challenges.
Now it all needed to work from the player’s couch.
The earliest changes were those that were easiest for Niantic to make on-the-fly, but they had dramatic impacts on the way the game actually works.
Some of the changes:
Doubling the players “radius” for interacting with in-game gyms, landmarks that players can temporarily take over for their in-game team, earning occupants a bit of in-game currency based on how long they maintain control. This change let more gym battles happen from the couch.
Increasing spawn points, generally upping the number of Pokémon you could find at home dramatically.
Increasing “incense” effectiveness, which allowed players to use a premium item to encourage even more Pokémon to pop up at home. Niantic phased this change out in October, then quietly reintroduced it in late November. Incense would also last twice as long, making it cheaper for players to use.
Allowing steps taken indoors (read: on treadmills) to count toward in-game distance challenges.
Players would no longer need to walk long distances to earn entry into the online player-versus-player battle system.
Your “buddy” Pokémon (a specially designated Pokémon that you can level up Tamagotchi-style for bonus perks) would now bring you more gifts of items you’d need to play. Pre-pandemic, getting these items meant wandering to the nearby “Pokéstop” landmarks.
By twisting some knobs and tweaking variables, Pokémon GO became much easier to play without leaving the house — but, importantly, these changes avoided anything that might break the game while being just as easy to reverse once it became safe to do so.
GO Fest goes virtual
Like this, just … online. Image Credits: Greg Kumparak
Thrown by Niantic every year since 2017, GO Fest is meant to be an ultra-concentrated version of the Pokémon GO experience. Thousands of players cram into one park, coming together to tackle challenges and capture previously unreleased Pokémon.
How Niantic evolved Pokémon GO for the year no one could go anywhere