DataDownload: The precarious nature of Facebook’s metaverse A weekly summary of all things Media, Data, Emerging Tech View this email in your browser
Have you ever played a VR game, or gone to a meeting or conference in VR, or taken a vacation in VR… chances are you’ve done one of those or more. So, when Slate asked the question “If Social Media Can Be Unsafe for Kids, What Happens in VR?” it’s worth considering. Now, to be fair, they said the same thing about Rock and Roll music, and TV, and violent video games and movies. You decide for yourself where to draw the line.
MIT Technology Review reports on a new swizzle on the AI Fake-Face generator — which now can tell you what the underlying (real) person was. This is the kind of transparency we can deeply endorse.
We’ve got a clip from Jon Stewart with David Remnick of The New Yorker. It’s topical, but you decide for yourself if Steward is back, or just visiting. And the BBC gives us a peek into people’s dreams.
Last week’s Summit was so chock full of interesting ideas and projects that we’re still thinking through everything we learned. Future Imperfect? Certainly seems so. Oh, and one parting gift — Captain Kirk in Space. For so many reasons, 90 years old perhaps the most amazing.
Have a great week.
The NYC Media Lab Must-Read If Social Media Can Be Unsafe for Kids, What Happens in VR?
On the heels of WSJ’s excellent investigative series, The Facebook Files, and the company suspending plans for “Instagram Kids”… and whistleblower Frances Haugen, one has to wonder at Facebook’s “bad shit happening” multiplier as it transitions into a metaverse company, “controlling a global, immersive, shared 3D digital space, including integrated social VR applications.” Slate suggests that now, in the words of Helen Lovejoy, is a good time to think of the children.
“At the moment, most companies, including Facebook, still recommend (not require) that children under 13 not use their VR products. At the same time, VR adoption remains slow, relative to other, less cumbersome technologies. This means there is already a need, but also a window of opportunity to learn from, and to not repeat, the mistakes of the past — the viral hate, misogyny, and racism, harassment, security breaches, voter interference, disinformation, algorithmic biases, conspiracy theories, child exploitation.”
Slate / 20 min read
Aside from tackling grand science challenges — and knocking researchers off their feet with SOTA protein folding prediction models and weather forecasting systems — DeepMind is taking on classical computer science. The neural algorithmic reasoning (NAR) team at DeepMind — composed of senior researchers Charles Blundell and Petar Veličković — is trying to develop a deep learning system that generates algorithms themselves.
Some of deep learning’s most impressive milestones have been narrow-domain solutions; it’s difficult to re-apply these models in different contexts. Meanwhile, an “algorithm might give you a great solution, but there’s really no way of saying whether this human-devised heuristic is actually the best way of looking at it.”
That’s why Blundell and Veličković are trying to abstract away the heuristics bit by swapping the algorithm with a neural network: “The key idea in NAR is to replace the algorithm with a neural network, typically a graph neural network (GNN)… this high dimensional GNN neural executor needs to actually imitate the algorithm at hand. For this reason, NAR also includes an abstract pipeline where that neural network is pre-trained, typically using lots of synthetic data.”
VentureBeat / 14 min read Read More Tech+Media Google Cloud Will Now Show Its Users Their Carbon Footprint in the Cloud
Google is adding free carbon footprint reports to its cloud services, with the idea of streamlining the process of reporting the environmental impact of companies’ cloud usage. According to Jenn Bennett, who leads Google Cloud’s data and technology strategy for sustainability in the Office of the CTO:
“Customers can monitor their cloud emissions over time by project, by product and by region, empowering IT teams and developers with metrics that help them reduce their carbon footprint. Digital infrastructure emissions are really just one part of their environmental footprint, but accounting for carbon emissions is necessary to measure progress against the carbon reduction targets that they all have.”
TechCrunch / 1 min read Read More AI Fake-Face Generators Can Be Rewound to Reveal the Real Faces They Trained On
The cool thing about This Person Does Not Exist has always been that, at the press of the refresh button, you’d get a totally unique person that actually never existed. But in This Person (Probably) Exists, researchers demonstrated that while the generated faces are fake, they can act as pointers, many resembling the people they were trained on. To uncover this flaw, University of Caen Normandy researchers used an adversarial technique called a membership attack:
“These attacks typically take advantage of subtle differences between the way a model treats data it was trained on… and unseen data. For example, a model might identify a previously unseen image accurately, but with slightly less confidence than one it was trained on. A second, attacking model can learn to spot such tells in the first model’s behavior and use them to predict when certain data, such as a photo, is in the training set or not. Such attacks can lead to serious security leaks. For example, finding out that someone’s medical data was used to train a model associated with a disease might reveal that this person has that disease.”
MIT Technology Review / 5 min read Read More The Largely Untold Story of How One Guy in California Keeps the World’s Computers Running on the Right Time Zone.
For the most part, timezones have been painful thorns in the sides of developers, or the type of boring terminal-based stuff that belongs in the computer equivalent of the Mesozoic era for the rest of us. But like the Tedium newsletter keeps pointing out to us, the apparently boring usually turns out to be surprisingly fascinating, or downright weird.
The internet’s official timezone database falls somewhere in between. In one developer’s dive down the timezone rabbit hole, we learn that all Linux and Mac-based computer pull their timezone data from the “zoneinfo database,” that the community maintaining this database is at war with itself, and that it has the audacity to tell countries that they cannot prematurely end DST. And that the person at the helm of a database that is relied upon by hundreds of millions of users is Paul Eggert, a UCLA computer scientist. Hence the xkcd strip below.
OneZero / 14 min read
Read More What We’re Watching Jon Stewart on Trump, Cancel Culture, and Optimism
“Jon Stewart talks with David Remnick about his thoughts on cancel culture — and how, even in a precarious time, he maintains hope.”
The New Yorker (YouTube) / 6 min watch
“Bond music sounds as cool as the man himself. We show you exactly how a few musical motifs shaped one of the most enduring film music legacies in history, from a 60s jazz orchestra to Billie Eilish. Listen to this episode’s companion playlist on Spotify.”
Spotify / 26 min listen
Listen Now Virtual Events Free Event: Developing Products For Emerging Countries
Date: October 19, 9:30AD EDT
“Bruno has over 11 years of experience in empowering people with easy access to products and services. Currently, he is a Senior Product Manager of Payments at Amazon.” Register Here. A Deeper Look How Scientists Learned to Enter People’s Dreams
The science of dreaming is pretty shaky. Most of us don’t have perfect recall of what occured during our slumber, or usually we don’t remember anything at all. Yet people’s recollections (and physiological data) is a large part of what researchers rely on today. “If that’s all we have to go on, then building a solid science of dreaming will be difficult,”says Ken Paller, a psychologist and dream researcher at Northwestern University.
Paller and his team have been exploiting lucid dreams to sidestep this game of mental telephone. Communicating with lucid dreamers is easier — they’re aware and are able to indicate to researchers that they’re in a lucid dream. That’s been known for decades, but Paller and colleague Karen Konkoly went a step further than past experiments:
“After participants indicated they were in a lucid dream, the scientists gave them basic maths questions, such as ‘eight minus six’, which the participants answered successfully using eye movements, according to a code agreed earlier (for instance, in this case, the answer ‘two’ was communicated by a left-right, left-right eye movement).”
BBC Science Focus Magazine / 10 min read