27 April 2020

Google’s much improved Pixel Buds are finally here


The original Pixel Buds weren’t very good. No way around it. Here’s a thing I wrote about them in a review titled “A disappointing debut for Google’s Pixel Buds“: As recently as a couple of years ago, they would have been a contender for the most compelling Bluetooth headphones on the market. But given the strides much of the competition has made, they mostly land with a dull thud.”

And we weren’t alone. Google’s first attempt at wireless earbuds were met with a pretty resounding “meh,” when they arrived in 2017. It’s probably an understatement to suggest that the company went back to the drawing board on this one. The line required a rethink from the ground up.

It took another two and a half years to deliver their successor. And Google seemingly sought to wipe the slate clean entirely, even going so far as not listing a “2” in the name. The new Pixel Buds are simply Pixel Buds. Anything else you remember with that name was clearly a figment of your own imagination.

Those original Pixel Buds that definitely didn’t exist already felt outdated when they hit the market. And while a clean slate was certainly required, Google didn’t do itself any favors by waiting that long. The landscape for wireless earbuds has grown by leaps and bounds in that time. The market has been saturated and the products feel more of a necessity than a luxury.

Six months after their introduction at a Pixel event in New York, the Buds are finally available for purchase in the U.S. — in Clearly White, at least. The other, more fun colors — Oh So Orange, Almost Black and Quite Mint — are not yet on the market. A minor quibble for those who have waited this long for a decent pair of Google headphones.

Color issues aside, I’m pretty into the design language here. It feels fresh in a way most earbuds don’t — the case in particular. It would have been easy to knock-off Apple or Samsung or any number of competitors, but the new Pixel Buds manage to pull off a fresh aesthetic built on top of the same basic concept of charging case that’s essentially universal across the board, at this point.

I actually prefer the matte black to the AirPod gloss. It’s better to look at and feels nice to the touch. Jury’s still out on how easily it will scratch. Full disclosure, I haven’t really left my apartment since the Buds arrived — because, well, life. The case is ovular — a flattened egg, if you will. The top of the case opens with an easy flip. There’s a black accent running around the lid, easily showing where to stick your thumb.

The case is fairly long in relation to the Buds themselves, owing, one imagines, to the size of the battery. All told, the Buds should get 24 hours with the case. There’s a USB-C port on the bottom (they’re wirelessly chargable, too) and a pairing button on the rear. The charging light flips on when open — white for full, orange for low battery.

Flipping the case open with the Buds in will also trigger a pairing dialog box on Pixel phones and other handsets running Android 6.0 and up. It’s a super simple pairing process — one akin to what you’ll get with AirPods on iOS. And once the headphones are registered to you, the box will pop up with the info on your other devices.

The Buds themselves are also aesthetically distinct from most of the competition. They feature a round button surface sporting a small, engraved Google “G.” The surface gives you space for the touch controls, which are as follows:

  • Tap to play/pause media, answer calls
  • Double tap to skip track, end/reject call, stop the Assistant
  • Triple tap to rewind/go to previous track
  • Swipe forward to increase volume
  • Swipe backward to decrease volume

The Buds felt good in my ears with the default medium tips. There are a larger and smaller pair in the box, as well, so you can play around to get a better fit. They’ve been in for the better part of four hours and my ears feel fine — not something I can say with every pair of earbuds I’ve tested. They’re not too large or heavy, so they don’t pull on or press the ear. There’s also a small, removable silicone wing at the top to keep them in place.

The battery on the buds is a bit lacking. After the aforementioned amount of time, I just got a low battery notification on the right bud. Curiously, they’ve run down at different rates. The right is at 14%, the left at 34%. Time to stick them back in the case for a recharge.

The sound is decent. Not the best sounding pair I’ve tried and certainly not the worst. I’d say they’re pretty middle of the pack in terms of the price point. If audio is (understandably) you’re biggest concern, I’d recommend opting for a pricier model from Sony, Sennheiser or Apple’s AirPods Pro. There’s no active noise canceling here, either. The “Hey Google” microphone array works as advertised whether activated by voice or a long press with a finger. The connection was mostly solid. I was able to keep the music playing while walking into another room, though I did hit a few rough patches here and there.

At $179, the new Pixel Buds are priced close to the middle of the pack. That feels about right. The models are a big upgrade over their disappointing predecessors, but are still a pretty middle of the road choice for Android users.


Read Full Article

TikTok launches Donation Stickers, allowing creators to fundraise for coronavirus relief efforts


TikTok is making it easier for creators and their fans to donate to favorite charities amid the coronavirus pandemic. The company today announced the launch of a new, interactive feature, Donation Stickers, that creators can use on their videos and live streams in order to raise funds for favorite charities directly in the TikTok app. At launch, these stickers will work to support charitable partners including CDC Foundation, James Beard Foundation, Meals on Wheels, MusiCares, National PTA, National Restaurant Association Educational Foundation, No Kid Hungry, and The Actors Fund.

The stickers work like any other, in terms of being added to a video or a TikTok LIVE stream. However, when a user taps on the sticker, they’ll be guided to a pop-up window where they can make a donation to the charity the creator is fundraising for — without ever having to leave the TikTok app.

The donations themselves are being powered by charitable fundraising platform, Tiltify, which handles the payment processing for the donation transactions. Tiltify has experience with donation features embedded in live streams, having previously worked with the Twitch platform on various initiatives.

TikTok says the charitable organizations it partnered with for the launch of the feature includes those whose current missions to support vulnerable groups that are also reflective of TikTok communities.

The app today is among those being adopted by doctors, nurses and other health care workers. These medical professionals see TikTok a means of of connecting younger users with credible health information about the coronavirus outbreak and COVID-19 at a time when conspiracy theories and bogus “cures” are being marketed across social media, and even the president is making dangerous off-the-cuff remarks not backed by science.

In addition, many of the other causes supported by the Donation Stickers align with communities hit hard by coronavirus shutdowns — like actors, musicians, educators and restaurant workers, for example.

The company says it will also match donations raised through the Donation Stickers until May 27th. The hashtag #doubleyourimpact will be automatically added to videos and live streams that use the stickers, as a result.

“During this time of uncertainty, our community has come together and given back in countless ways, from applauding health care workers to sharing inspirational messages on how to stay safe and happy at home to making original coronavirus songs to spread positive messages,” wrote TikTok U.S. Head of Product, Sean Kim, in an announcement about the stickers’ launch. “We’ve been impressed and heartened by the selfless steps our community has taken to help each other, and now we’re excited to be able to give our users another way to make a positive impact.”

The addition of the stickers is one of several ways TikTok has been involved in coronavirus relief efforts. The company earlier this month pledged $250 million to support front-line workers, educators and local communities affected by the COVID-19 pandemic. It also provided an additional $125 million in advertising credits to public health organizations and businesses looking to rebuild.

The donation-matching through the new stickers will come from this $250 million fund. As part of the previously announced Community Relief Fund, TikTok is donating $4 million to No Kid Hungry and Meals on Wheels.

On May 5th through May 9th, TikTok is also hosting a week of TikTok LIVE streams focused on fundraising as a part of the “Happy At Home: #OneCommunity” nightly event.


Read Full Article

Crisis support for the world, one text away | Nancy Lublin

Crisis support for the world, one text away | Nancy Lublin

What if we could help people in crisis anytime, anywhere with a simple text message? That's the idea behind Crisis Text Line, a free 24-hour service that connects people in need with trained, volunteer crisis counselors -- "strangers helping strangers around the world, like a giant global love machine," as cofounder and CEO Nancy Lublin puts it. Learn more about their big plans to expand to four new languages, providing a third of the globe with crucial, life-saving support. (This ambitious plan is a part of the Audacious Project, TED's initiative to inspire and fund global change.)

Click the above link to download the TED talk.

YouTube and Tribeca announce We Are One, a 10-day online film festival


With COVID-19 making it unsafe to watch movies in crowded theaters, not to mention traveling for the red-carpet glamor of a film festival, many festival organizers have been looking at online alternatives.

So today, YouTube and Tribeca Enterprises (the organization behind New York City’s Tribeca Film Festival) are announcing a new event called We Are One: A Global Film Festival.

It’s not simply an online replacement for Tribeca, but aims to be a truly global event. The 10-day digital film festival will include programing curated by representatives from most of the major film festivals around the world.

We Are One kicks off on May 29 and is supposed to benefit the World Health Organization’s COVID-19 Solidarity Response Fund, as well as local relief providers.

“We are proud to join with our partner festivals to spotlight truly extraordinary films and talent, allowing audiences to experience both the nuances of storytelling from around the world and the artistic personalities of each festival,” said Pierre Lescure and Thierry Frémaux of the Cannes Film Festival (which will not be taking place this year in its “original form”) in a statement.

While the programming featured during the 10-day event hasn’t been announced yet, participating festivals include:

  • the Annecy International Animation Film Festival
  • Berlin International Film Festival
  • BFI London Film Festival
  • Cannes Film Festival
  • Guadalajara International Film Festival
  • International Film Festival & Awards Macao
  • Jerusalem Film Festival
  • Mumbai Film Festival
  • Karlovy Vary International Film Festival
  • Locarno Film Festival
  • Marrakech International Film Festival
  • New York Film Festival
  • San Sebastian International Film Festival
  • Sarajevo Film Festival
  • Sundance Film Festival
  • Sydney Film Festival
  • Tokyo International Film Festival
  • Toronto International Film Festival
  • Tribeca Film Festival
  • Venice Film Festival


Read Full Article

The 5 Best Laptop Cooling Mats


If you are using your personal laptop in your home office, you may be concerned about keeping it in top condition. When it overheats, this spells disaster for your motherboard or even your hard drive.

Nobody wants to deal with a computer in meltdown, so keeping your laptop cool is essential. Here are some of the best laptop cooling mats to keep your computer frosty.

1. Targus Portable Lightweight Chill Mat

Targus Portable Lightweight Chill Mat Targus Portable Lightweight Chill Mat Buy Now On Amazon $39.99

Targus has long been manufacturing tech peripherals that won’t tear a hole in your wallet. Their Portable Lightweight Chill Mat is one such device. The comapny’s design means it is ideal for placing on your lap. Pretty much where you’d expect your laptop to sit. You can still use it on your desk, though, if you like.

The dimensions of the Targus are 15 x 1.00 x 11.75 inches, so it doesn’t have the smallest of footprints. However, as you are using it on your knee, that won’t matter too much. Neoprene covers the plastic frame, keeping the unit’s weight at a minimum. The bottom of the Targus mat has four rubber feet to keep the mat in place on a hard, smooth surface. The top has four rubber bumpers to hold your laptop in position.

One advantage of this cooling mat is its angled construction. This relieves the stress on your wrists as you type, making it more comfortable to use. A USB powers the cooling mat’s two large fans, which plugs into the computer itself. You can also use a USB hub if your ports are a bit on the scarce side.

2. Cooler Master NotePal XL Laptop Cooling Pad

Cooler Master NotePal XL Laptop Cooling Pad Cooler Master NotePal XL Laptop Cooling Pad Buy Now On Amazon $96.00

Cooler Master’s NotePal XL Laptop Cooling Pad is quite the futuristic-looking accessory. That said, it’ll also keep your laptop cool too. The frame is made from durable plastic to ensure the mat stays lightweight. This is due to the metal mesh surface upon which you place the laptop itself. The metal mesh provides maximum airflow to the underside of your laptop. This is the area most prone to overheating.

A single large fan takes care of the cooling. With a diameter of 230mm, the fan should provide cooling to almost all of your laptop’s base. Great news, because that means it will work with a wide range of different laptops. Overall, the Cooler Master measures in at 78 x 117 x 140 mm, so it won’t take up too much prime real estate on your desk. Like the Targus, this pad is angled for your comfort. You can increase the angle using the adjustable legs hidden in the base.

A blue X-shaped LED light adds a nice visual touch to the Cooler Master option. However, one of the main highlights is the ability to use the cooling mat as a USB hub. A control panel to the rear boasts three USB 2.0 for output, a mini-USB for power in, and a micro-USB for external power. You can even adjust the fan speed if you wish. Cooler Master claims that the fan is silent, so it shouldn’t interrupt your work, either.

3. TopMate C5 Gaming Laptop Cooler

TopMate C5 Gaming Laptop Cooler TopMate C5 Gaming Laptop Cooler Buy Now On Amazon $42.99

Gaming laptops are arguably more prone to overheating than a general, everyday work laptop. This may be because the laptop is being pushed harder, performance-wise. If you use your laptop for gaming as well as working from home, then you need a dedicated cooling mat. The TopMate C5 Gaming Laptop Cooler is just the ticket. Designed specifically with gaming laptops in mind, it will ensure your valuable computer doesn’t suffer an untimely death.

Like the Cooler Master mat, the TopMate C5 features snazzy blue LEDs to light up the cooling mat. This gives it a futuristic, arcade-like feel—perfect for you to incorporate it into your gaming rig. There are no less than five fans to cool your computer, providing excellent coverage to the underside of your laptop. There are four outer fans and one larger central fan for added cooling to your computer’s core.

You can control the fans using the small LCD screen at the front of the TopMate C5. This gives you six different fan speeds and three fan operation modes. You can customize the TopMate laptop cooler further, too. Its height is adjustable in five increments, giving you plenty of angling options. A hinged flap at the bottom pops up to provide a shelf for your laptop to rest against. This stops it from taking a nose-dive towards the floor, giving you peace of mind.

4. Thermaltake Massive 20 RGB Laptop Cooling Pad

Thermaltake Massive 20 RGB Laptop Cooling Pad Thermaltake Massive 20 RGB Laptop Cooling Pad Buy Now On Amazon $49.99

Style and substance come in spades with Thermaltake’s Massive 20 RGB Laptop Cooling Pad. If you want your laptop mat to make an impression, then this is the one for you. Like the TopMate, the Thermaltake Massive 20 is intended for gaming and other workload-intensive laptops. The edges of the frame feature full 256-color LED lights, so it will look awesome sitting on your desk. Avoid laptop burnout while you jazz up your office space!

With a footprint measuring 18.5 x 14 x 1.5 inches, it isn’t the smallest on the market. However, if you’re planning on popping your prized hi-spec laptop on it, you want to make sure it can take the added weight and heat. A large metal mesh offers plenty of room for a 19-inch gaming laptop, with the airflow optimized by the supporting platter. A 200mm silent fan ensures cool air is constantly blowing against your laptop’s base.

The rear of the Thermaltake laptop cooling pad features a small control panel. With this, you can control the light, color, and fan speed. Further customization comes in the form of the angle adjustment. Three adjustable height settings can angle the Thermaltake cooling pad at 3 degrees, 9 degrees, and 13 degrees. Not only does this make it more ergonomic, but it can also increase airflow to the base of your laptop.

5. Belkin CoolSpot Laptop Cooling Pad

Belkin CoolSpot Laptop Cooling Pad Belkin CoolSpot Laptop Cooling Pad Buy Now On Amazon $91.76

Belkin’s CoolSpot Laptop Cooling Pad is designed with less intensive use in mind. If your laptop is a bit on the older side, it might struggle to perform under the increased load of working from home. As it is your line of communication with colleagues and clients, you don’t want it to suddenly self-combust. You need a cooling pad, but not an all-singing-all-dancing one.

The CoolSpot boasts a nice, flowing design. This ensures the device is ergonomic and comfortable to use. It is only 11.7 x 1.8 x 11.4 inches, so it is small and lightweight. This makes it perfect for an intermediate user. It also boasts Belkin’s unique vortex fan design. This single fan combines with the CoolSpot’s Airflow Wave design to ensure maximum cooling efficiency.

Your laptop only sits on rubber pads at the top and bottom of the CoolSpot, ensuring peak airflow below your computer. The slim design means that you can slip it into your laptop bag, making this cooling pad effortlessly portable. USB power and quiet operation ensure you can use it in the university library or your favorite coffee shop.

The Best Laptop Cooling Pad for You

The range of laptop cooling pads above should give you enough options to make the right choice when you buy. If you are working from home, then you may feel that you need to deck out the rest of your office.

If so, check out our essential home office accessories and get your workspace functioning as it should.

Read the full article: The 5 Best Laptop Cooling Mats


Read Full Article

How Will COVID-19 Contact Tracing Apps Help Break Coronavirus Lockdown


The COVID-19 pandemic is forcing governments to investigate citizens like never before. A confirmed COVID-19 case can transmit the coronavirus onto countless other people. Tracing any individual that encounters the deadly coronavirus could help stop further transmission, isolating the spread, and potentially assisting with the lifting of lockdown measures.

As you might expect, there is some opposition to an app that traces and matches your location. Even if it has a net positive effect, is tracing your location a breach of privacy?

So, how do COVID-19 contact tracing apps work? And can a contact tracing app protect your privacy?

What Is a COVID-19 Contact Tracing App?

A COVID-19 contact tracing app is a tool that governments and healthcare professionals can use to trace the movements of an individual who has coronavirus.

A contact tracing app will capture the locations a person has visited during the time they are suspected of having the coronavirus. After establishing a list of locations, the contact tracing app can trace any other smartphones that were in the vicinity of the individual during that period.

An app can send out messages to affected citizens automatically, inform those in high-risk groups to seek medical advice, and more. The sooner a person knows they were in contact with someone carrying COVID-19, the quicker they can begin to mediate their behavior, be that self-isolating or seeking treatment.

COVID-19 contact tracing app development is, understandably, in overdrive. There are various projects around the world attempting to create a contact tracing solution. The difficulty is balancing the need to locate, trace, and inform, versus the very real issue of protecting the privacy of the individual.

How Does a COVID-19 Contact Tracing App Work?

There are multiple coronavirus contact tracing apps in development. At the time of writing, at least 30 countries are developing or already implementing COVID-19 contact tracing apps. The apps use several different approaches and privacy frameworks.

There are two main approaches to coronavirus contact tracing.

  • Centralized Contact Tracing: A centralized approach to COVID-19 contact tracing uses the network location of a smartphone, rather than an individual app. The centralized approach eliminates the need to download an app, decreasing the number of citizens that would avoid contact tracing apps. Centralizing contact tracing through network location creates a significant privacy issue, however.
  • Decentralized/Privacy-Focused Contact Tracing: Privacy-focused contact tracing methods (also known as privacy-preserving contact tracing) do exist. These apps use a different range of smartphone technologies for monitoring and tracing. Several privacy-focused contact tracing frameworks use Bluetooth Low Energy (BLE) connections to approximate a user’s location and proximity to other smartphones. Furthermore, Apple and Google are working in partnership to develop a contact tracing project (more on this below).

It isn’t just privacy advocates sounding their concerns regarding coronavirus contact tracing apps. The scale of contact tracing is forcing everyone to consider how such apps will protect user privacy.

Furthermore, the decentralized, privacy-focused options are not without fault. One idea mooted had support from hundreds of respected academics, privacy advocates, and security researchers. Yet, once the project published its project, many pulled their support, citing a lack of oversight and that the project would not protect privacy as first indicated.

How Can a Contact Tracing App Protect Privacy and Remain Effective?

That is the question each contact tracing protocol development team is attempting to solve. At the time of writing, five privacy-preserving contact tracing (PPCT) protocols are in development. Three PPCT protocols are garnering more interest than other options—though not all for the right reasons.

Pan-European Privacy-Preserving Proximity Tracing (PEPP-PT)

PEPP-PT was one of the first privacy-preserving contact tracing apps to begin picking up development speed.

PEPP-PT uses BLE to track and log users in close proximity to a user. The protocol then sends the data to a centralized server for contact processing, where potentially infected users are contacted. If a user is a confirmed coronavirus case, they receive a request to upload their contact log for analysis.

Although PEPP-PT claims strong privacy credentials, the project received criticism regarding transparency on the functionality of the protocol, as well as why the PEPP-PT had released no open-source code for scrutiny.

When PEPP-PT did publish in-depth details of how the protocol works, including the use of centralized servers, researchers and academics associated with the project began switching support to the DP-3T project (see below). Over 300 academics and researchers removed their support from the project on April 20, 2020.

“Such apps can otherwise be repurposed to enable unwarranted discrimination and surveillance,” said a joint letter signed by academics in over 26 countries. “It is crucial that citizens trust the applications in order to produce sufficient uptake to make a difference in tackling the crisis. It is vital that, in coming out of the current crisis, we do not create a tool that enables large-scale data collection on the population, either now or at a later time.”

Decentralized Privacy-Preserving Proximity Tracing (DP-3T)

Decentralized Privacy-Preserving Proximity Tracing is an open-source contact tracing project that uses BLE to track and log users. Like PEPP-PT, the contact and location information uploads to a server.

However, DP-3T uses an “ephemeral, pseudo-random ID” to represent the user. It also uses the same pseudo-random ID to document any interaction with another user. The DP-3T app broadcasts the temporary random ID to other smartphones. Any smartphones in the same proximity also receive a temporary random ID.

If the user becomes a COVID-19 patient, they can upload their local app location data. The users remains anonymized through the pseudo-random ID. The app detects the potential for contact with other users and sends a message accordingly (also using the pseudo-random ID of the other users).

Although the DP-3T protocol still uses a central server, it has several integrated privacy protection features. For instance, the app shares no information with any healthcare service until the user uploads their location data. This prevents the abuse of personal data as no single entity receives a tranche of data, especially data not meant for a specific organization or otherwise.

The server itself cannot uncover an individual identity on the network, because each user keeps their data local until the point of upload.

Finally, the DP-3T project has confirmed that it will dismantle the app at the end of the COVID-19 pandemic. Importantly, any “data on the server is removed after 14 days.”

Several countries are implementing DP-3T contact tracing apps to assist with stopping the spread of COVID-19.

Google/Apple PPCT Project

Google and Apple are working on a coronavirus contact tracing app that would utilize their smartphone operating systems (Android and iOS, respectively). As the two companies control the smartphone operating system market, the tech giants hold a unique place in the battle against COVID-19.

The “Gapple” PPCT project uses a similar system to DP-3T, using BLE interactions to trace users. The log uses randomized identifiers to protect the privacy of all parties. The identifiers change every 15 minutes to anonymize the data further.

Data is held locally for 14 days. If the user doesn’t receive a contact tracing message during that time, the app deletes the data, including any identifiers.

Issues with Using Bluetooth Low Energy for COVID-19 Contact Tracing

As you have seen, each solution proposes using Bluetooth Low Energy for coronavirus contact tracing apps.

Bluetooth and its successor, Bluetooth Low Energy, are ubiquitous throughout most countries around the world. However, an estimated 2 billion mobile phones around the world do not use BLE. A further 1.5 billion use legacy phones that do not run a modern mobile operating system.

Exacerbating the issue is the fact that most of the people in that bracket are more vulnerable to COVID-19, be that due to age, location, or income demographic.

Another BLE issue is the technology itself. Bluetooth Low Energy can transmit over distances of 10 to 30 meters, depending on the device. The commonly accepted social distancing advice is to remain 2 meters apart from each other. But if your phone can ping someone up to 30 meters away, there will be false positives.

Due to the way the contact tracing apps work, a single false positive could cause a cascade of false-positive messages through the alleged connections of that user.

Furthermore, coverage is key to the efficacy of any privacy-preserving contact tracing app, BLE, or not.

In the UK, researchers from the University of Oxford estimate that at least 80 percent of the smartphone owning population must install the contact tracing app to reach a reasonable level of coverage. The figure equates to around 56 percent of the UK population.

Which leads to another issue. If someone doesn’t want to use a COVID-19 contact tracing app, they just won’t download it. A similar system developed in Singapore had an uptake of just 12 percent. That’s not nearly enough to create an effective contact tracing system.

Will Contact Tracing Help Stop COVID-19?

There are issues regarding the implementation of coronavirus contact tracing apps. However, a slow consensus is building, recognizing that some form of social distancing management is going to have to be in place in order to return to “real life.”

The onus is on building contact tracing apps that protect the privacy of the users. As you might expect, if there is a hiccup with a contact tracing app, it can expose private user data.

For example, an early iteration of a contact tracing app in South Korea broadcast the personal data of coronavirus cases while alerting those who may have come into contact. The developers quickly patched the flaw from the contact tracing app. Still, those fears surrounding privacy remain strong, especially in countries yet to begin widespread contact tracing testing.

In the US, there is a strong indifference to COVID-19 contact tracing apps, with many respondents to a recent Pew Research study indicating little faith in the system.

pew research us citizens distrust contact tracing

The delicate balance between preserving privacy and working to protect public health is fraught with potential pitfalls. In Israel, the government proposes to use counterterrorism laws to instigate network-level tracking of all devices. That’s beyond the pale and a level of intrusion that most citizens will never accept.

But if it means society and the economy can begin functioning as normal, some form of contact tracing is inevitable, at least in the short term.

Will Coronavirus Contact Tracing Destroy Privacy?

The idea of supporting another smartphone tracing method goes against our inbuilt desire for privacy. On the Joe Rogan Experience podcast, Edward Snowden explains in great detail how your smartphone is already the number one tracking tool in the world.

A series of protocols that extend tracking to every phone in your vicinity is another surveillance step.

On the other hand, COVID-19 is affecting millions of people around the world. The implementation of DP-3T stores data locally to stop any other party engaging with your location information until you catch the coronavirus.

If the government wanted to track you, they would already be doing it. An app that could save lives is a worthwhile endeavor in the short term, especially as many countries begin to relax lockdown regulations and begin to wonder what a 2nd COVID-19 peak might entail.

Part of the difficulty healthcare professionals face is misinformation regarding COVID-19. Check out these sites for trustworthy up to date coronavirus news. Another issue facing everyone is the surge in coronavirus-related phishing attacks. Here’s how to spot a COVID-19 phishing attempt and how to stay safe.

Read the full article: How Will COVID-19 Contact Tracing Apps Help Break Coronavirus Lockdown


Read Full Article

How to See and Download Your Netflix Viewing History


netflix-viewing-history

Many people don’t know that Netflix keeps a running log of your viewing history (activity). While slightly buried on Netflix’s account page options, your Netflix viewing history can be invaluable.

So the next time you’re struggling to remember what you watched last month, take a look and download it for yourself. Simply follow these steps to access your Netflix viewing history.

How to See Your Viewing Activity

netflix activity settings

Before doing anything else, you need to log in to Netflix from your browser. Just remember, if you use the Netflix app to access your account settings, you’ll still be redirected to use your mobile browser. So save yourself a step by doing it on your computer.

Once you’ve logged into Netflix on your browser, look for your profile icon in the upper-right corner of your screen and follow the steps below:

  1. Hover over your profile image and click on Account once it appears.
  2. Scroll down to Profile & Parental controls.
  3. Click on the profile you want to see the viewing activity for.
  4. Scroll down and click on viewing activity.
  5. Your results will now be displayed. You can repeat this process on any other profile.

How to Download Your Viewing Activity

download netflix viewing activity

Now that you’ve accessed your activity, you’ll see a few options. At the top of the page, you can report a problem with any of your recent episodes. You can also choose to hide individual episodes from your viewing history.

However, to download your Netflix viewing history you need to go to the bottom of the page where you can choose to show more, hide all, or download all. Select download all, and you’ll automatically download your Netflix viewing history as a .csv file.

How to Open Your Downloaded Netflix Viewing History

netflix viewing history excel

If you’re unfamiliar with .csv files, there are two easy ways to open them. You can use software such as Microsoft Excel or OpenOffice Calc. Or you can also stick with your browser and use Google Documents.

Once you know what Google Docs is and how to use it, you simply have to navigate a few menus.

  1. Go to Google Docs and sign in, if you haven’t already.
  2. Under Start a new document, click on Blank.
  3. Select File and then select Open.
  4. Click the Upload tab.
  5. Drag the .csv file there, or select a file from your device, navigate to it, and click open.
  6. The spreadsheet will automatically open in Google Sheets.

Watch More Netflix to Expand Your Viewing Activity

Now that you’re able to see and download your Netflix viewing history, you’ll always know what you have watched and when you watched it. It’s a great first step towards managing what you watch on Netflix. This also allows you to monitor what content other accounts have accessed as well.

In addition, you can use your Netflix viewing history to figure out what to watch with friends on Netflix. And for more information on that, check out our article explaining how to watch Netflix with friends far away.

Read the full article: How to See and Download Your Netflix Viewing History


Read Full Article

Germany ditches centralized approach to app for COVID-19 contacts tracing


Germany has U-turned on building a centralized COVID-19 contacts tracing app — and will instead adopt a decentralized architecture, Reuters reported Sunday, citing a joint statement by chancellery minister Helge Braun and health minister Jens Spahn.

In Europe in recent weeks, a battle has raged between different groups backing centralized vs decentralized infrastructure for apps being fast-tracked by governments which will use Bluetooth-based smartphone proximity as a proxy for infection risk — in the hopes of supporting the public health response to the coronavirus by automating some contacts tracing.

Centralized approaches that have been proposed in the region would see pseudonymized proximity data stored and processed on a server controlled by a national authority, such as a healthcare service. However concerns have been raised about allowing authorities to scoop up citizens’ social graph, with privacy experts warning of the risk of function creep and even state surveillance.

Decentralized contacts tracing infrastructure, by contrast, means ephemeral IDs are stored locally on device — and only uploaded with a user’s permission after a confirmed COVID-19 diagnosis. A relay server is used to broadcast infected IDs — enabling devices to locally compute if there’s a risk that requires notification. So social graph data is not centralized.

The change of tack by the German government marks a major blow to a homegrown standardization effort, called PEPP-PT, that had been aggressively backing centralization — while claiming to ‘preserve privacy’ on account of not tracking location data. It quickly scrambled to propose a centralized architecture for tracking coronavirus contacts, led by Germany’s Fraunhofer Institute, and claiming the German government as a major early backer, despite PEPP-PT later saying it would support decentralized protocols too.

As we reported earlier, the effort faced strident criticism from European privacy experts — including a group of academics developing a decentralized protocol called DP-3T — who argue p2p architecture is truly privacy preserving. Concerns were also raised about a lack of transparency around who is behind PEPP-PT and the protocols they claimed to support, with no code published for review.

The European Commission, meanwhile, has also recommended the use of decentralization technologies to help boost trust in such apps in order to encourage wider adoption.

EU parliamentarians have also warned regional governments against trying to centralize proximity data during the coronavirus crisis.

But it was Apple and Google jumping into the fray earlier this month by announcing joint support for decentralized contacts tracing that was the bigger blow — with no prospect of platform-level technical restrictions being lifted. iOS limits background access to Bluetooth for privacy and security reasons, so national apps that do not meet this decentralized standard won’t benefit from API support — and will likely be far less usable, draining battery and functioning only if actively running.

Nonetheless PEPP-PT told journalists just over a week ago that it was engaged in fruitful discussions with Apple and Google about making changes to their approach to accommodate centralized protocols.

Notably, the tech giants never confirmed that claim. They have only since doubled down on the principle of decentralization for the cross-platform API for public health apps — and system-wide contacts tracing which is due to launch next month.

At the time of writing PEPP-PT’s spokesman, Hans-Christian Boos, had not responded to a request for comment on the German government withdrawing support.

Boos previously claimed PEPP-PT had around 40 governments lining up to join the standard. However in recent days the momentum in Europe has been going in the other direction. A number of academic institutions that had initially backed PEPP-PT have also withdrawn support.

In a statement emailed to TechCrunch, the DP-3T project welcomed Germany’s U-turn. “DP-3T is very happy to see that Germany is adopting a decentralized approach to contact tracing and we look forward to its next steps implementing such a technique in a privacy preserving manner,” the group told us.

Berlin’s withdrawal leaves France and the UK the two main regional backers of centralized apps for coronavirus contacts tracing. And while the German U-turn is certainly a hammer blow for the centralized camp in Europe the French government appears solid in its support — at least for now.

France has been developing a centralized coronavirus contacts tracing protocol, called ROBERT, working with Germany’s Fraunhofer Institute and others.

In an opinion issued Sunday, France’s data protection watchdog, the CNIL, did not take active issue with centralizing pseudonymized proximity IDs — saying EU law does not in principle forbid such a system — although the watchdog emphasized the need to minimize the risk of individuals being re-identified.

It’s notable that France’s digital minister, Cédric O, has been applying high profile public pressure to Apple over Bluetooth restrictions — telling Bloomberg last week that Apple’s policy is a blocker to the virus tracker.

Yesterday O was also tweeting to defend the utility of the planned ‘Stop Covid’ app.

We reached out to France’s digital ministry for comment on Germany’s decision to switch to a decentralized approach but at the time of writing the department had not responded.

In a press release today the government highlights the CNIL view that its approach is compliant with data protection rules, and commits to publishing a data protection impact assessment ahead of launching the app.

If France presses ahead it’s not clear how the country will avoid its app being ignored or abandoned by smartphone users who find it irritating to use. (Although it’s worth noting that Google’s Android platform has a substantial marketshare in the market, with circa 80% vs 20% for iOS, per Kantar.)

A debate in the French parliament tomorrow is due to include discussion of contacts tracing apps.

We’ve also reached out to the UK’s NHSX — which has been developing a COVID-19 contacts tracing app for the UK market — and will update this report with any response.

In a blog post Friday the UK public healthcare unit’s digital transformation division said it’s “working with Apple and Google on their welcome support for tracing apps around the world”, a PR line that entirely sidesteps the controversy around centralized vs decentralized app infrastructures.

The UK has previously been reported to be planning to centralize proximity data — raising questions about the efficacy of its planned app too, given iOS restrictions on background access to Bluetooth.

“As part of our commitment to transparency, we will be publishing the key security and privacy designs alongside the source code so privacy experts can ‘look under the bonnet’ and help us ensure the security is absolutely world class,” the NHSX’s Matthew Gould and Dr Geraint Lewis added in the statement.


Read Full Article

WhatsApp’s new limit cuts virality of ‘highly forwarded’ messages by 70%


WhatsApp’s bid on cutting virality of messages circulating on its platform by introducing an additional limit earlier this month has already started to pay off.

The Facebook-owned service said on Monday that spread of “highly forwarded” messages sent on WhatsApp had dropped by 70% globally in weeks after introducing a new restriction earlier this month.

In one of the biggest changes to its core feature, WhatsApp said earlier this month that users on its platform can now send along frequently forwarded messages they receive to only one person or a group at a time, down from five. The restriction was rolled out globally to WhatsApp’s 2 billion users on April 7.

“We recently introduced a limit to sharing ‘highly forwarded messages’ to just one chat. Since putting into place this new limit, globally there has been a 70% reduction in the number of highly forwarded messages sent on WhatsApp,” a WhatsApp spokesperson told TechCrunch in a statement.

WhatsApp first introduced a similar limit in 2018, when it restricted users from forwarding a message to more than five people or groups at once. While announcing the new restriction earlier this month, WhatsApp said message forwards on its service had dropped by 25% globally in two years.

“This change is helping keep WhatsApp a place for personal and private conversations. WhatsApp is committed to doing our part to tackle viral messages,” the spokesperson said today.

The cut down on forwards should help WhatsApp assuage the scrutiny it is receiving in many countries, including India, its biggest market.

New Delhi asked WhatsApp and other messaging and social media firms last month to do more to control the viral hoaxes circulating on their platforms about coronavirus infection. This is the latest of several similar advisories India has sent to social media firms operating in the country,

WhatsApp said earlier this month that it had seen a “significant” surge in the “amount of forwarding” in recent weeks that “users have told us can feel overwhelming and can contribute to the spread of misinformation.”

In recent weeks, several users in India have circulated messages, often in good faith, that claimed that treatments had been found to battle coronavirus infection and that there were scientific explanations to back some of the community measures New Delhi had enforced such as asking people to make noise for five minutes or lighting candles and oil-lamps. Fact checkers said that none of these claims were factual.

WhatsApp and its parent firm, Facebook, have taken several efforts in recent months to help governments in many countries, including India, reach their citizens and share authoritative information about the coronavirus pandemic.


Read Full Article

Shine adds invoice insurance to its freelancer bank account


French startup Shine is adding a new option today. If you think there’s a chance that a client is not going to pay your next invoice, you can insure that invoice to avoid any bad surprise.

Shine is building a challenger bank for freelancers and small companies. It lets you send and receive money in a separate business account, pay with a MasterCard, create invoices and stay on top of administrative tasks.

It also helps you get started as the startup can fill out all administrative paperwork to register yourself as a freelancer. You also get notifications to remind you that you should pay your taxes and more. Starting accepting freelancing jobs can be confusing and Shine can help you with that.

Shine has a built-in invoicing tool. It lets you add a client and generate an invoice directly in the mobile app. After that, you can send a link to your client. You get a notification when your client opens the invoice. They can download a PDF and get your bank details to pay you.

And yet, many clients often wait until the last minute to pay an invoice. It can be a month or two after finishing a job, which means that they also forget about outstanding invoices.

In a few weeks, Shine users will be able to create an invoice and insure it before sending it. It costs you 2% of your total amount on your invoice. There’s no subscription fee, it’s a one-off process.

If your client hasn’t paid you after the due date, Shine will reach out to your client again to try to get the payment. If that doesn’t work, you can file a claim with the partner insurance company.

In that case, if the company is still operating, you get paid 100% of your invoice. If the company has collapsed, you get 90% back. (Of course, that’s without taking into account the 2% fees you already paid.)


Read Full Article

Google at ICLR 2020




This week marks the beginning of the 8th International Conference on Learning Representations (ICLR 2020), a fully virtual conference focused on how one can learn meaningful and useful representations of data for machine learning. ICLR offers conference and workshop tracks, both of which include invited talks along with oral and poster presentations of some of the latest research on deep learning, metric learning, kernel learning, compositional models, non-linear structured prediction and issues regarding non-convex optimization.

As a Diamond Sponsor of ICLR 2020, Google will have a strong virtual presence with over 80 publications accepted, in addition to participating on organizing committees and in workshops. If you have registered for ICLR 20202, we hope you'll watch our talks and learn about the projects and opportunities at Google that go into solving interesting problems for billions of people. You can also learn more about our research being presented at ICLR 2020 in the list below (Googlers highlighted in blue).

Officers and Board Members
Includes: Hugo LaRochelle, Samy Bengio, Tara Sainath

Organizing Committee
Includes: Kevin Swersky, Timnit Gebru

Area Chairs
Includes: Balaji Lakshminarayanan, Been Kim, Chelsea Finn, Dale Schuurmans, George Tucker, Honglak Lee, Hossein Mobahi, Jasper Snoek, Justin Gilmer, Katherine Heller, Manaal Faruqui, Michael Ryoo, Nicolas Le Roux, Sanmi Koyejo, Sergey Levine, Tara Sainath, Yann Dauphin, Anders Søgaard, David Duvenaud, Jamie Morgenstern, Qiang Liu

Publications
SEED RL: Scalable and Efficient Deep-RL with Accelerated Central Inference (see the blog post)
Lasse Espeholt, Raphaël Marinier, Piotr Stanczyk, Ke Wang, Marcin Michalski‎

Differentiable Reasoning Over a Virtual Knowledge Base
Bhuwan Dhingra, Manzil Zaheer, Vidhisha Balachandran, Graham Neubig, Ruslan Salakhutdinov, William W. Cohen

Dynamics-Aware Unsupervised Discovery of Skills
Archit Sharma, Shixiang Gu, Sergey Levine, Vikash Kumar, Karol Hausman

GenDICE: Generalized Offline Estimation of Stationary Values
Ruiyi Zhang, Bo Dai, Lihong Li, Dale Schuurmans

Mathematical Reasoning in Latent Space
Dennis Lee, Christian Szegedy, Markus N. Rabe, Kshitij Bansal, Sarah M. Loos

Your Classifier is Secretly an Energy Based Model and You Should Treat it Like One
Will Grathwohl, Kuan-Chieh Wang, Jorn-Henrik Jacobsen, David Duvenaud, Kevin Swersky, Mohammad Norouzi

Adjustable Real-time Style Transfer
Mohammad Babaeizadeh, Golnaz Ghiasi

Are Transformers Universal Approximators of Sequence-to-sequence Functions?
Chulhee Yun, Srinadh Bhojanapalli, Ankit Singh Rawat, Sashankc J. Reddi, Sanjiv Kumar

AssembleNet: Searching for Multi-Stream Neural Connectivity in Video Architectures
Michael S. Ryoo, AJ Piergiovanni, Mingxing Tan, Anelia Angelova

AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty
Dan Hendrycks, Norman Mu, Ekin D. Cubuk, Barret Zoph, Justin Gilmer, Balaji Lakshminarayanan

BatchEnsemble: an Alternative Approach to Efficient Ensemble and Lifelong Learning
Yeming Wen, Dustin Tran, Jimmy Ba

Black-box Off-policy Estimation for Infinite-Horizon Reinforcement Learning (see the blog post)
Ali Mousavi, Lihong Li, Qiang Liu, Dengyong Zhou

Can Gradient Clipping Mitigate Label Noise?
Aditya Krishna Menon, Ankit Singh Rawat, Sashank J. Reddi, Sanjiv Kumar

CAQL: Continuous Action Q-Learning
Moonkyung Ryu, Yinlam Chow, Ross Anderson, Christian Tjandraatmadja, Craig Boutilier

Chameleon: Adaptive Code Optimization for Expedited Deep Neural Network Compilation
Byung Hoon Ahn, Prannoy Pilligundla, Amir Yazdanbakhsh, Hadi Esmaeilzadeh

Coherent Gradients: An Approach to Understanding Generalization in Gradient Descent-based Optimization
Satrajit Chatterjee

Consistency Regularization for Generative Adversarial Networks
Han Zhang, Zizhao Zhang, Augustus Odena, Honglak Lee

Contrastive Representation Distillation
Yonglong Tian, Dilip Krishnan, Phillip Isola

Deep Audio Priors Emerge from Harmonic Convolutional Networks
Zhoutong Zhang, Yunyun Wang, Chuang Gan, Jiajun Wu, Joshua B. Tenenbaum, Antonio Torralba, William T. Freeman

Detecting and Diagnosing Adversarial Images with Class-Conditional Capsule Reconstructions
Yao Qin, Nicholas Frosst, Sara Sabour, Colin Raffel, Garrison Cottrell, Geoffrey Hinton

Detecting Extrapolation with Local Ensembles
David Madras, James Atwood, Alexander D'Amour

Disentangling Factors of Variations Using Few Labels
Francesco Locatello, Michael Tschannen, Stefan Bauer, Gunnar Rätsch, Bernhard Schölkopf, Olivier Bachem

Distance-Based Learning from Errors for Confidence Calibration
Chen Xing, Sercan Ö. Arik, Zizhao Zhang, Tomas Pfister

ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators (see the blog post)
Kevin Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning

ES-MAML: Simple Hessian-Free Meta Learning (see the blog post)
Xingyou Song, Yuxiang Yang, Krzysztof Choromanski, Aldo Pacchiano, Wenbo Gao, Yunhao Tang

Exploration in Reinforcement Learning with Deep Covering Options
Yuu Jinnai, Jee Won Park, Marlos C. Machado, George Konidaris

Extreme Tensoring for Low-Memory Preconditioning
Xinyi Chen, Naman Agarwal, Elad Hazan, Cyril Zhang, Yi Zhang

Fantastic Generalization Measures and Where to Find Them
Yiding Jiang, Behnam Neyshabur, Hossein Mobahi, Dilip Krishnan, Samy Bengio

Generalization Bounds for Deep Convolutional Neural Networks
Philip M. Long, Hanie Sedghi

Generalized Convolutional Forest Networks for Domain Generalization and Visual Recognition
Jongbin Ryu, GiTaek Kwon, Ming-Hsuan Yang, Jongwoo Lim

Generative Models for Effective ML on Private, Decentralized Datasets
Sean Augenstein, H. Brendan McMahan, Daniel Ramage, Swaroop Ramaswamy, Peter Kairouz, Mingqing Chen, Rajiv Mathews, Blaise Aguera y Arcas

Generative Ratio Matching Networks
Akash Srivastava, Kai Xu, Michael U. Gutmann, Charles Sutton

Global Relational Models of Source Code
Vincent J. Hellendoorn, Petros Maniatis, Rishabh Singh, Charles Sutton, David Bieber

Hierarchical Foresight: Self-Supervised Learning of Long-Horizon Tasks via Visual Subgoal Generation
Suraj Nair, Chelsea Finn

Identity Crisis: Memorization and Generalization Under Extreme Overparameterization
Chiyuan Zhang, Samy Bengio, Moritz Hardt, Michael C. Mozer, Yoram Singer

Imitation Learning via Off-Policy Distribution Matching
Ilya Kostrikov, Ofir Nachum, Jonathan Tompson

Language GANs Falling Short
Massimo Caccia, Lucas Caccia, William Fedus, Hugo Larochelle, Joëlle Pineau, Laurent Charlin

Large Batch Optimization for Deep Learning: Training BERT in 76 Minutes
Yang You, Jing Li, Sashank Reddi, Jonathan Hseu, Sanjiv Kumar, Srinadh Bhojanapalli, Xiaodan Song, James Demmel, Kurt Keutzer, Cho-Jui Hsieh

Learning Execution through Neural Code Fusion
Zhan Shi, Kevin Swersky, Daniel Tarlow, Parthasarathy Ranganathan, Milad Hashemi

Learning Heuristics for Quantified Boolean Formulas through Reinforcement Learning
Gil Lederman, Markus N. Rabe, Edward A. Lee, Sanjit A. Seshia

Learning to Learn by Zeroth-Order Oracle
Yangjun Ruan, Yuanhao Xiong, Sashank Reddi, Sanjiv Kumar, Cho-Jui Hsieh

Learning to Represent Programs with Property Signatures
Augustus Odena, Charles Sutton

MACER: Attack-free and Scalable Robust Training via Maximizing Certified Radius
Runtian Zhai, Chen Dan, Di He, Huan Zhang, Boqing Gong, Pradeep Ravikumar, Cho-Jui Hsieh, Liwei Wang

Measuring Compositional Generalization: A Comprehensive Method on Realistic Data
Daniel Keysers, Nathanael Schärli, Nathan Scales, Hylke Buisman, Daniel Furrer, Sergii Kashubin, Nikola Momchev, Danila Sinopalnikov, Lukasz Stafiniak, Tibor Tihon, Dmitry Tsarkov, Xiao Wang, Marc van Zee, Olivier Bousquet

Meta Reinforcement Learning with Autonomous Inference of Subtask Dependencies
Sungryull Sohn, Hyunjae Woo, Jongwook Choi, Honglak Lee

Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples
Eleni Triantafillou, Tyler Zhu, Vincent Dumoulin, Pascal Lamblin, Utku Evci, Kelvin Xu, Ross Goroshin, Carles Gelada, Kevin Swersky, Pierre-Antoine Manzagol, Hugo Larochelle

Model-based Reinforcement Learning for Biological Sequence Design
Christof Angermueller, David Dohan, David Belanger, Ramya Deshpande, Kevin Murphy, Lucy Colwell

Network Randomization: A Simple Technique for Generalization in Deep Reinforcement Learning
Kimin Lee, Kibok Lee, Jinwoo Shin, Honglak Lee

Observational Overfitting in Reinforcement Learning
Xingyou Song, Yiding Jiang, Stephen Tu, Behnam Neyshabur, Yilun Du

On Bonus-based Exploration Methods In The Arcade Learning Environment
Adrien Ali Taiga, William Fedus, Marlos C. Machado, Aaron Courville, Marc G. Bellemare

On Identifiability in Transformers
Gino Brunner, Yang Liu, Damian Pascual, Oliver Richter, Massimiliano Ciaramita, Roger Wattenhofer

On Mutual Information Maximization for Representation Learning
Michael Tschannen, Josip Djolonga, Paul K. Rubenstein, Sylvain Gelly, Mario Lucic

On the Global Convergence of Training Deep Linear ResNets
Difan Zou, Philip M. Long, Quanquan Gu

Phase Transitions for the Information Bottleneck in Representation Learning
Tailin Wu, Ian Fischer

Pre-training Tasks for Embedding-based Large-scale Retrieval
Wei-Cheng Chang, Felix X. Yu, Yin-Wen Chang, Yiming Yang, Sanjiv Kumar

Prediction, Consistency, Curvature: Representation Learning for Locally-Linear Control
Nir Levine, Yinlam Chow, Rui Shu, Ang Li, Mohammad Ghavamzadeh, Hung Bui

Provable Benefit of Orthogonal Initialization in Optimizing Deep Linear Networks
Wei Hu, Lechao Xiao, Jeffrey Pennington

Rapid Learning or Feature Reuse? Towards Understanding the Effectiveness of MAML
Aniruddh Raghu, Maithra Raghu, Samy Bengio, Oriol Vinyals

Reinforced Genetic Algorithm Learning for Optimizing Computation Graphs
Aditya Paliwal, Felix Gimeno, Vinod Nair, Yujia Li, Miles Lubin, Pushmeet Kohli, Oriol Vinyals

ReMixMatch: Semi-Supervised Learning with Distribution Alignment and Augmentation Anchoring
David Berthelot, Nicholas Carlini, Ekin D. Cubuk, Alex Kurakin, Han Zhang, Colin Raffel, Kihyuk Sohn

Scalable Model Compression by Entropy Penalized Reparameterization
Deniz Oktay, Johannes Ballé, Saurabh Singh, Abhinav Shrivastava

Scalable Neural Methods for Reasoning With a Symbolic Knowledge Base
William W. Cohen, Haitian Sun, R. Alex Hofer, Matthew Siegler

Semi-Supervised Generative Modeling for Controllable Speech Synthesis
Raza Habib, Soroosh Mariooryad, Matt Shannon, Eric Battenberg, RJ Skerry-Ryan, Daisy Stanton, David Kao, Tom Bagby

Span Recovery for Deep Neural Networks with Applications to Input Obfuscation
Rajesh Jayaram, David Woodruff, Qiuyi Zhang

Thieves on Sesame Street! Model Extraction of BERT-based APIs
Kalpesh Krishna, Gaurav Singh Tomar, Ankur P. Parikh, Nicolas Papernot, Mohit Iyyer

Thinking While Moving: Deep Reinforcement Learning with Concurrent Control
Ted Xiao, Eric Jang, Dmitry Kalashnikov, Sergey Levine, Julian Ibarz, Karol Hausman, Alexander Herzog

VideoFlow: A Conditional Flow-Based Model for Stochastic Video Generation
Manoj Kumar, Mohammad Babaeizadeh, Dumitru Erhan, Chelsea Finn, Sergey Levine, Laurent Dinh, Durk Kingma

Watch, Try, Learn: Meta-Learning from Demonstrations and Rewards
Allan Zhou, Eric Jang, Daniel Kappler, Alex Herzog, Mohi Khansari, Paul Wohlhart, Yunfei Bai, Mrinal Kalakrishnan, Sergey Levine, Chelsea Finn

Weakly Supervised Disentanglement with Guarantees
Rui Shu, Yining Chen, Abhishek Kumar, Stefano Ermon, Ben Poole

You Only Train Once: Loss-Conditional Training of Deep Networks
Alexey Dosovitskiy, Josip Djolonga

A Mutual Information Maximization Perspective of Language Representation Learning
Lingpeng Kong, Cyprien de Masson d’Autume, Wang Ling, Lei Yu, Zihang Dai, Dani Yogatama

ALBERT: A Lite BERT for Self-supervised Learning of Language Representations (see the blog post)
Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut

Asymptotics of Wide Networks from Feynman Diagrams
Ethan Dyer, Guy Gur-Ari

DDSP: Differentiable Digital Signal Processing
Jesse Engel, Lamtharn Hantrakul, Chenjie Gu, Adam Roberts

Doubly Robust Bias Reduction in Infinite Horizon Off-Policy Estimation
Ziyang Tang, Yihao Feng, Lihong Li, Dengyong Zhou, Qiang Liu

Dream to Control: Learning Behaviors by Latent Imagination (see the blog post)
Danijar Hafner, Timothy Lillicrap, Jimmy Ba, Mohammad Norouzi

Emergent Tool Use From Multi-Agent Autocurricula
Bowen Baker, Ingmar Kanitscheider, Todor Markov, Yi Wu, Glenn Powell, Bob McGrew, Igor Mordatch

Gradientless Descent: High-Dimensional Zeroth-Order Optimization
Daniel Golovin, John Karro, Greg Kochanski, Chansoo Lee, Xingyou Song, Qiuyi (Richard) Zhang

HOPPITY: Learning Graph Transformations to Detect and Fix Bugs in Programs
Elizabeth Dinella, Hanjun Dai, Ziyang Li, Mayur Naik, Le Song, Ke Wang

Learning to Plan in High Dimensions via Neural Exploration-Exploitation Trees
Binghong Chen, Bo Dai, Qinjie Lin, Guo Ye, Han Liu, Le Song

Model Based Reinforcement Learning for Atari (see the blog post)
Łukasz Kaiser, Mohammad Babaeizadeh, Piotr Miłos, Błazej Osinski, Roy H. Campbell, Konrad Czechowski, Dumitru Erhan, Chelsea Finn, Piotr Kozakowski, Sergey Levine, Afroz Mohiuddin, Ryan Sepassi, George Tucker, Henryk Michalewski

Neural Symbolic Reader: Scalable Integration of Distributed and Symbolic Representations for Reading Comprehension
Xinyun Chen, Chen Liang, Adams Wei Yu, Denny Zhou, Dawn Song, Quoc V. Le

SUMO: Unbiased Estimation of Log Marginal Probability for Latent Variable Models
Yucen Luo, Alex Beatson, Mohammad Norouzi, Jun Zhu, David Duvenaud, Ryan P. Adams, Ricky T. Q. Chen

Measuring the Reliability of Reinforcement Learning Algorithms
Stephanie C.Y. Chan, Samuel Fishman, John Canny, Anoop Korattikara, Sergio Guadarrama

Meta-Learning without Memorization
Mingzhang Yin, George Tucker, Mingyuan Zhou, Sergey Levine, Chelsea Finn

Neural Tangents: Fast and Easy Infinite Neural Networks in Python (see the blog post)
Roman Novak, Lechao Xiao, Jiri Hron, Jaehoon Lee, Alexander A. Alemi, Jascha Sohl-Dickstein, Samuel S. Schoenholz

Scaling Autoregressive Video Models
Dirk Weissenborn, Oscar Täckström, Jakob Uszkoreit

The Intriguing Role of Module Criticality in the Generalization of Deep Networks
Niladri Chatterji, Behnam Neyshabur, Hanie Sedghi

Reformer: The Efficient Transformer (see the blog post)
Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya

Workshops
Computer Vision for Global Challenges
Organizing Committee: Ernest Mwebaze
Advisory Committee: Timnit Gebru, John Quinn

Practical ML for Developing Countries: Learning under limited/low resource scenarios
Organizing Committee: Nyalleng Moorosi, Timnit Gebru
Program Committee: Pablo Samuel Castro, Samy Bengio
Keynote Speaker: Karmel Allison

Tackling Climate Change with Machine Learning
Organizing Committee: Moustapha Cisse
Co-Organizer: Natasha Jaques
Program Committee: John C. Platt, Kevin McCloskey, Natasha Jaques
Advisor and Panel: John C. Platt

Towards Trustworthy ML: Rethinking Security and Privacy for ML
Organizing Committee: Nicholas Carlini, Nicolas Papernot
Program Committee: Shuang Song