23 February 2019

Oppo announces 5G and 10x lossless zoom handsets


Saturday afternoon is a rough time for a press conference — particularly with the official kickoff of Mobile World Congress still a few days away. That said, there are certain advantages to being an early bird. Chief among them is the ability to claim firsts — namely having the first 5G handset of the show.

That might not mean a lot in the grand scheme of things, but in a week that’s expected to be dominated by 5G announcements, it’s a way to stand out from the crowd. Of course, like the rest of the promised 5G handsets we’ve heard about so far — with the noble exception of Samsung’s — details are still pretty scarce

What we do know is that the handset — along with so many others set to be announced this week — will be powered by Qualcomm’s Snapdragon 855. Fitting, given that we can almost certainly expect some 5G news out of the chipmaker this week. Oppo also says the device will be on display on the show floor this week — actually firing it up and experiencing those next generation speeds in person, however, is a different thing entirely.

Another bit of news out of the event is the promise of 10x lossless zoom (16mm-160mm) for the company’s next flagship. If its works as advertised that’s a nice little distinguisher from the competition — though 10x zoom likely isn’t a day to day feature for most smartphone users. That device is due out at some point in Q2. 


Read Full Article

Opera Touch Now Lets You Block Cookie Dialog Boxes


Opera’s cookie dialog blocker is now available for Opera Touch on Android and iOS. The cookie dialog blocker means you’ll never have to see another cookie dialog box. Which have become the bane of everyone’s browsing experience since GDPR was introduced.

Opera launched Opera Touch on iOS in October 2018, bringing one-handed browsing to Apple’s operating system. The company has since made dozens of improvements to Opera Touch, including a new cookie dialog blocker, and more besides.

Block Those Annoying GDPR Cookie Prompts

Opera’s cookie dialog blocker does exactly what the name suggests. Which is let you block those annoying cookie dialog boxes that pop up when you visit a website for the first time. Or, if you’re prone to clearing your browsing data, every time you visit a website, period.

Cookie dialog boxes have become a major annoyance since the introduction of the General Data Protection Regulation (GDPR). GDPR is intended to help people stay in control of their data, but it’s just meant we all have to tick more boxes.

Opera’s cookie dialog blocker solves this problem. Once enabled, the cookie dialog blocker will hide these prompts from view. By default this means you’ll accept all cookies, but you can also reject them. However, this may prevent some websites from working.

To enable Opera’s cookie dialog blocker on Opera Touch, click the three dot menu in the top-right of the browser, click Settings, and then click Block Cookie Dialogs. You can then move the sliders across to suit your own needs, either accepting or rejecting cookies.

Download: Opera Touch for Android or iOS

24 Ways Opera Has Improved Opera Touch

While the cookie dialog blocker is the newest Opera Touch feature, the company is keen to point out other improvements it has made. On the Opera Blog, Opera details 24 improvements, including a Private Mode, a Dark Theme, and configurable search engines.

If you haven’t yet tried the one-handed browser it would be worth revisiting our article discussing the launch of Opera Touch. And if you’re a fan of all things Opera and want to catch a glimpse of its plans for the future, check out our first look at Opera Neon.

Read the full article: Opera Touch Now Lets You Block Cookie Dialog Boxes


Read Full Article

Share YouTube Videos Quickly With These 5 Minimal Tools


YouTube has one-third of the internet in its claws. More than 1.9 billion people enjoy YouTube every month. You can bet that translates to a lot of watched videos. But how does the world share YouTube videos? Maybe you do it from your mobile or use Facebook. It does matter because there are several ways to share YouTube videos. Some are more efficient than others.

Let’s look at five quick techniques to share YouTube videos and enjoy them with our friends.

Riv.yt: Share Distraction Free Videos Instantly

Share YouTube Videos Without Distractions

YouTube can get in its own way. One interesting video leads to another. This can be a problem if you want everyone to give your video undivided attention. Or you want a kid-friendly way to share videos without the nonsense of comments and advertisements. Riv.yt has the answer.

Riv.yt is a minimalist way of sharing YouTube videos without the distractions. Use the simple search to find a video of your choice and then choose a design for the landing page. Share the generated link which carries your video enveloped in a beautiful landing page.

The only thing missing right now is a dark mode or a dark theme.

YouTube Time: Share From a Specific Time in a YouTube Video

Link to a Specific Part of a YouTube Video

YouTube itself gives you an easy way to start a video from a specific time. Skip to the part you want and right click on the video. Select Copy video URL at current time and then share it. But YouTube Time doesn’t make it more difficult. YouTube Time makes it precise to the second.

Paste your YouTube URL, then enter the exact start time in the box for minute and seconds. You can preview the cut video or generate the link to share it via any other medium. You can also use this URL as an embed in your blog or in a forum. The video will skip to the part you set it up for.

YouTube Time is a clean, minimal site. And this is just one of the important YouTube URL tricks you should keep in mind to create or share “ad free” videos by skipping around them.

Kapwing: Resize and Share Videos Socially

Trim and share YouTube videos

You have your video but it’s the wrong size for Instagram or the Facebook Newsfeed. Video resizing tools can get complicated. Enter Kapwing which just needs a URL and the choice of the platform you want to optimize for. The minimal UI does the rest.

Crop to the perfect ratio or even add padding and fit the whole video in the frame so that it doesn’t look awkward or get cut off. Kapwing is  useful when you want to post a portrait video on Instagram.

GIFit (Chrome): Share YouTube Clips as GIFs

Make animated GIFs from YouTube videos.

Clip a bit of footage from a YouTube video and share it as a GIF. Now, where have we heard that before? Everywhere… because GIFs are all the rage for the last few years and they have almost taken the place of our emotions. GIPHY and even Google Search give you animated GIFs to choose from. But YouTube is a monster source of memes that will continue to give you the ones no one has shared before.

GIFit is a handy Chrome extension that helps you clip a part from a YouTube video and share it as an animated GIF. Install the Chrome extension and head to YouTube. Select a video and you can spot the GIFIt button next to the Closed Caption button on the toolbar. Set a start time, end time, and the quality you want. Hit the GIfIt button to create your GIF.

Save and share it anywhere. It works well on the fly. But if you want to add a slogan or create a professional social media post, then look at a few serious video editors to create memes, montages, or timelapses. Or try Hashcut which we have featured before.

Download: GIFit for Chrome (Free)

vdNote: Share Videos and Timestamped Notes

vdNote

We have covered VdNote before, but it deserves a recall because of its sheer utility. The minimal interface is a plus. vdNote is one of the tools you should have in your arsenal if you use YouTube for learning. With vdNote, you can enter comments refer to the exact moment in the video you were watching. When you share the videos, others can use the comments to understand the context.

Go to vdNote and paste in a link to a YouTube video you want to annotate & start watching. At the right point, type a note in the text field. Notice that the video pauses to give you time to finish your thoughts.

You can share your video and the connected notes with their timestamps via email or Twitter.

Share YouTube Videos with Context

The secret sauce of viral videos isn’t exponential sharing. It is the meaning we enjoy while watching the video. But sharing is easy. Creating your own original videos and then sharing it can be more meaningful.

How about switching from watching videos to making them yourself? YouTube allows you to make all kinds of videos and you will be surprised at the variety in the successful ones.

Read the full article: Share YouTube Videos Quickly With These 5 Minimal Tools


Read Full Article

YouTube demonetizes anti-vaccination videos


YouTube will demonetize channels that promote anti-vaccination views, after a report by BuzzFeed News found ads, including from health companies, running before anti-vax videos. The platform will also place a new information panel that links to the Wikipedia entry on “vaccine hesitancy” before anti-vax videos. Information panels (part of YouTube’s efforts to combat misinformation) about the measles, mumps, and rubella (MMR) vaccine had already appeared in front of anti-vaccination videos that mentioned it.

In a statement to BuzzFeed News, a YouTube spokesperson said “we have strict policies that govern what videos we allow ads to appear on, and videos that promote anti-vaccination content are a violation of those policies. We enforce these policies vigorously, and if we find a video that violates them, we immediately take action and remove ads.”

This is the second issue this week that has prompted YouTube advertisers to suspend their ads BuzzFeed News’ initial report on Feb. 20 came as several major advertisers, including Nestle and Epic Games, said they were pausing ads after YouTube creator Matt Watson revealed how the platform’s recommendation algorithm was being exploited by what he described as a “soft-core pedophilia ring.”

BuzzFeed News found that the top search results for queries about vaccine safety were usually from legitimate sources, like hospitals, but then YouTube’s Up Next algorithm would often recommend anti-vaccination videos. Ads, which are placed by YouTube’s advertising algorithm, appeared in front of many of those videos. YouTube also told BuzzFeed it would implement changes to its Up Next algorithm to prevent the spread of anti-vax videos.

Outbreaks of measles throughout the United States and in other countries have prompted scrutiny into the role of social media and tech companies, including Facebook and Google, in spreading misinformation.

Advertisers contacted by BuzzFeed News who said they will take action to prevent their ads from running in front of anti-vax videos include Nomad Health, Retail Me Not, Grammarly, Brilliant Earth, CWCBExpo, XTIVIA, and SolarWinds. Vitacost told BuzzFeed News that it had already pulled ads after the child exploitation issues became known.

Anto-vax channels now demonetized include VAXXED TV, LarryCook333, and iHealthTube.


Read Full Article

The 5 Best Sunlight Lamps for Winter Depression and Light Therapy

Twitter co-founder Ev Williams to step down from the company’s board


Ev Williams, a co-founder of Twitter and the social media business’s former chief executive officer, is stepping down from its board of directors effective at the end of the month, according to documents submitted to the U.S. Securities and Exchange Commission on Friday, first reported by CNBC.

In a series of tweets, Williams addressed the news.

“I’m very lucky to have served on the board for 12 years (ever since there was a board),” he wrote. “It’s been overwhelmingly interesting, educational—and, at times, challenging… Thank you, and for starting this crazy company with me—and continuing to make it better and better. And to my fellow board members, new and old—some of the most thoughtful people I’ve ever known.”

Williams, the founder and CEO of online publishing platform Medium and co-founder and partner at Obvious Ventures, served as Twitter’s chief executive from 2008 to 2010 following Jack Dorsey’s, Twitter’s current CEO, original stint as CEO. Williams was succeeded by Dick Costolo, who after a five-year stint at the helm, relinquished the throne back to Dorsey.

Twitter’s board of directors includes Bret Taylor, president and chief product officer at Salesforce; Debra Lee, president and CEO of BET Networks; and executive chairman Omid Kordestani.

Twitter’s stock closed up 3 percent Friday, trading at nearly $32 a piece for a market cap of north of $24 billion.


Read Full Article

Apple confirms its plans to close retail stores in the patent troll-favored Eastern District of Texas


Apple has confirmed its plans to close retail stores in the Eastern District of Texas – a move that will allow the company to better protect itself from patent infringement lawsuits, according to the Apple news site MacRumors which broke the news of the stores’ closures. Apple says that the impacted retail employees will be offered new jobs with the company as a result of these changes.

The company will shut down its Apple Willow Bend store in Plano, Texas as well as its Apple Stonebriar store in Frisco, Texas, MacRumors reported, and Apple confirmed. These stores will permanently close up shop on Friday, April 12. Customers in the region will instead be served by a new Apple store located at the Galleria Dallas Shopping Mall, which is expected to open April 13.

Apple did not comment on the stores’ dates of closure or the new store’s opening.

However, it’s common for Apple to leave little downtown during retail stores transitions – though most closures are related to renovations or other reasons that aren’t about trying to escape patent lawsuits.

The Eastern District of Texas had become a popular place for patent trolls to file their lawsuits, though a more recent Supreme Court ruling has attempted to crack down on the practice. The court ruled that patent holders could no longer choose where to file.

Apple has had to make big payouts to patent trolls in recent years: $625.6 million to patent holding firm VirnetX in 2016 over protocol patents; VirnetX won $368 million from Apple in 2013; and more recently $502.6 million over four communication patents.

VirnetX tends to be referred to as a “patent troll” because it makes most of its revenue by suing tech companies. In addition to Apple, it sued Microsoft over patents in Skype and has been in litigation with Cisco. Its cases and subsequent wins are often held up as another example of how patent law in the U.S. is in need of reform.

The Apple store closures could have had a notable impact on area jobs, had Apple not offered new positions to its retail staff.

Apple today employs 1,000 people in the Dallas-Fort Worth area, which has been an increase of 33 percent in the past five years.

The company also recently invested almost $30 million in its Dallas area stores.

Outside the metro, Apple is also investing in Texas with its $1 billion for the new campus in Austin, which will accommodate an additional 5,000 employees on top of the 6,200 already in the area.

A rep for Apple confirmed the stores’ closures in a statement, but wouldn’t comment on the company’s reasoning:

“We’re making a major investment in our stores in Texas, including significant upgrades to NorthPark Center, Southlake and Knox Street. With a new Dallas store coming to the Dallas Galleria this April, we’ve made the decision to consolidate stores and close Apple Stonebriar and Apple Willow Bend. All employees from those stores will be offered positions at the new Dallas store or other Apple locations.”

 


Read Full Article

Pinstagram? Instagram code reveals Public Collections feature


Instagram is threatening to attack Pinterest just as it files to go public the same way the Facebook-owned app did to Snapchat. Code buried in Instagram for Android shows the company has prototyped an option to create public “Collections” to which multiple users can contribute. Instagram launched private Collections two years ago to let you Save and organize your favorite feed posts. But by allowing users to make Collections public, Instagram would become a direct competitor to Pinterest.

Instagram public Collections could spark a new medium of content curation. People could use the feature to bundle together their favorite memes, travel destinations, fashion items, or art. That could cut down on unconsented content stealing that’s caused backlash against meme “curators” like F*ckJerry by giving an alternative to screenshotting and reposting other people’s stuff. Instead of just representing yourself with your own content, you could express your identity through the things you love — even if you didn’t photograph them yourself. And if that sounds familiar, you’ll understand why this could be problematic for Pinterest’s upcoming $12 billion IPO.

The “Make Collection Public” option was discovered by frequent TechCrunch tipster and reverse engineering specialist Jane Machun Wong. It’s not available to the public, but from the Instagram for Android code, she was able to generate a screenshot of the prototype. It shows the ability to toggle on public visibility for a Collection, and tag contributors who can also add to the Collection. Previously, Collections was always a private, solo feature for organizing your bookmarks gathered through the Instagaram Save feature Instagram launched in late 2016.

Instagram told TechCrunch “we’re not testing this” which is its standard response to press inquiries about products that aren’t available to any public users, but that are in internal development. It could be a while until Instagram does start experimenting publicly with the feature and longer before a launch, and the company could always scrap the option. But it’s a sensible way to give users more to do and share on Instagram, and the prototype gives insight into the app’s strategy. Facebook launched its own Pinterest-style shareable Sets in 2017 and launched sharable Collections in December.

Currently there’s nothing in the Instagram code about users being able to follow each other’s Collections, but that would seem like a logical and powerful next step. Instagrammers can already follow hashtags to see new posts with them routed to their feed. Offering a similar way to follow Collections could turn people into star curators rather than star creators without the need to rip off anyone’s content. Speaking of infuencers, Wong also spotted Instagram prototyping IGTV picture-in-picture so you could keep watching a long-form video after closing the app and navigating the rest of your phone.

Instagram lets users Save posts which can then be organized into Collections

Public Collections could fuel Instagram’s commerce strategy that Mark Zuckerberg recently said would be a big part of the roadmap. Instagram already has a personalized Shopping feed in Explore, and The Verge’s Casey Newton reported last year that Instagram was working on a dedicated shopping app. It’s easy to imagine fashionistas, magazines, and brands sharing Collections of their favorite buyable items.

It’s worth remembering that Instagram launched its copycat of Snapchat Stories just six months before Snap went public. As we predicted, that reduced Snapchat’s growth rate by 88 percent. Two years later, Snapchat isn’t growing at all, and its share price is at just a third of its peak. With over 1 billion monthly and 500 million daily users, Instagram is four times the size of Pinterest. Instagram loyalists might find it’s easier to use the ‘good enough’ public Collections feature where they already have a social graph than try to build a following from scratch on Pinterest.


Read Full Article

Pinstagram? Instagram code reveals Public Collections feature


Instagram is threatening to attack Pinterest just as it files to go public the same way the Facebook-owned app did to Snapchat. Code buried in Instagram for Android shows the company has prototyped an option to create public “Collections” to which multiple users can contribute. Instagram launched private Collections two years ago to let you Save and organize your favorite feed posts. But by allowing users to make Collections public, Instagram would become a direct competitor to Pinterest.

Instagram public Collections could spark a new medium of content curation. People could use the feature to bundle together their favorite memes, travel destinations, fashion items, or art. That could cut down on unconsented content stealing that’s caused backlash against meme “curators” like F*ckJerry by giving an alternative to screenshotting and reposting other people’s stuff. Instead of just representing yourself with your own content, you could express your identity through the things you love — even if you didn’t photograph them yourself. And if that sounds familiar, you’ll understand why this could be problematic for Pinterest’s upcoming $12 billion IPO.

The “Make Collection Public” option was discovered by frequent TechCrunch tipster and reverse engineering specialist Jane Machun Wong. It’s not available to the public, but from the Instagram for Android code, she was able to generate a screenshot of the prototype. It shows the ability to toggle on public visibility for a Collection, and tag contributors who can also add to the Collection. Previously, Collections was always a private, solo feature for organizing your bookmarks gathered through the Instagaram Save feature Instagram launched in late 2016.

Instagram told TechCrunch “we’re not testing this” which is its standard response to press inquiries about products that aren’t available to any public users, but that are in internal development. It could be a while until Instagram does start experimenting publicly with the feature and longer before a launch, and the company could always scrap the option. But it’s a sensible way to give users more to do and share on Instagram, and the prototype gives insight into the app’s strategy. Facebook launched its own Pinterest-style shareable Sets in 2017 and launched sharable Collections in December.

Currently there’s nothing in the Instagram code about users being able to follow each other’s Collections, but that would seem like a logical and powerful next step. Instagrammers can already follow hashtags to see new posts with them routed to their feed. Offering a similar way to follow Collections could turn people into star curators rather than star creators without the need to rip off anyone’s content. Speaking of infuencers, Wong also spotted Instagram prototyping IGTV picture-in-picture so you could keep watching a long-form video after closing the app and navigating the rest of your phone.

Instagram lets users Save posts which can then be organized into Collections

Public Collections could fuel Instagram’s commerce strategy that Mark Zuckerberg recently said would be a big part of the roadmap. Instagram already has a personalized Shopping feed in Explore, and The Verge’s Casey Newton reported last year that Instagram was working on a dedicated shopping app. It’s easy to imagine fashionistas, magazines, and brands sharing Collections of their favorite buyable items.

It’s worth remembering that Instagram launched its copycat of Snapchat Stories just six months before Snap went public. As we predicted, that reduced Snapchat’s growth rate by 88 percent. Two years later, Snapchat isn’t growing at all, and its share price is at just a third of its peak. With over 1 billion monthly and 500 million daily users, Instagram is four times the size of Pinterest. Instagram loyalists might find it’s easier to use the ‘good enough’ public Collections feature where they already have a social graph than try to build a following from scratch on Pinterest.


Read Full Article

Sebastian Thrun initiates aggressive plan to transform Udacity


“I’m a fighter. I believe in our people, I believe in our mission, and I believe that it should exist and must exist.”

Sebastian Thrun is talking animatedly about Udacity, the $1 billion online education startup that he co-founded nearly eight years ago. His tone is buoyant and hopeful. He’s encouraged, he says over an occasionally crackly phone call, about the progress the company has made in such a short time. There’s even a new interim COO, former HP and GE executive Lalit Singh, who joined just days ago to help Thrun execute this newly formed strategy.

That wasn’t the case four weeks ago.

In a lengthy email, obtained by TechCrunch, Thrun lobbed an impassioned missive to the entire company, which specializes in “nanodegrees” on a range of technical subjects that include AI, deep learning, digital marketing, VR and computer vision.

It was, at times, raw, personal and heartfelt, with Thrun accepting blame for missteps or admitting he was sleeping less than four hours a night; in other spots the email felt like a pep talk delivered by a coach, encouraging his team by noting their spirit and tenacity. There were moments when he exhibited frustration for the company’s timidness, declaring “our plans are ridden of fear, of trepidation, we truly suck!” And moments just as conciliatory, where he noted that “I know every one of you wants to double down on student success. I love this about us.”

Thrun has sent spirited emails before. Insiders say it’s not uncommon and that as a mission-driven guy he often calls on employees to take risks and be creative. But this one stood out for its underlying message.

If there was a theme in the email, it was an existential one: We must act, and act now or face annihilation.

“It was a rallying cry, to be honest,” Thrun told TechCrunch. “When I wrote this email, I really wanted to wake up people to the fact that our trajectory was not long-term tenable.”

“I can tell you that I woke up the troops, that is absolutely sure,” he said later. “Whether my strategy is sound, only time can tell.”

Udacity founder Sebastian Thrun speaks onstage during TechCrunch Disrupt SF 2017 at Pier 48 on September 19, 2017 in San Francisco, California. (Photo by Steve Jennings/Getty Images for TechCrunch)

Thrun said the past month has been transformative for the company. “It was a tough moment when I had to look at the business, look at the financials, look at the people in the company,” Thrun said, adding, “And the people in the company are amazing. I really believe in them, and I believe that they’re all behind the mission.”

A tough year

Part of Udacity’s struggles were borne out of its last funding round in 2015, when it raised $105 million and became a unicorn. That round and the valuation set high expectations for growth and revenue.

But the company started hitting those targets and 2017 became a breakout year.

After a booming 2017 — with revenue growing 40 percent year-over-year thanks to some popular programs like its self-driving car and deep learning nanodegrees — the following year fizzled. Its consumer business began to shrink, and while the production quality of its educational videos increased, the volume slowed.

“In 2018, we didn’t have a single a blockbuster,” Thrun said. “There’s nothing you can point to and say, ‘Wow, Udacity had a blockbuster.’ “

By comparison, the self-driving car engineering nanodegree not only was a hit, it produced a successful new company. Udacity vice president Oliver Cameron spun out an autonomous vehicle company called Voyage.

Udacity CEO Vishal Makhijani left in October and Thrun stepped in. He took over as chief executive and the head of content on an interim basis. Thrun, who founded X, Google’s moonshot factory, is also CEO of Kitty Hawk Corp., a flying-car startup.

His first impression upon his return was a company that had grown too quickly and was burdened by its own self-inflicted red tape. Staff reductions soon followed. About 130 people were laid off and other open positions were left vacant, Thrun said.

Udacity now has 350 full-time employees and another 200 full-time contractors. The company also has about 1,000 people contracted as graders or reviewers.

An emphasis, when I rejoined, was to cut complexity and focus the company on the things that are working,” he said. 

One area where Udacity seemed to excel had also created an impediment. The quality of Udacity’s video production resulted in Hollywood-quality programming, Thrun said. But that created a bottleneck in the amount of educational content Udacity could produce.

Udacity’s content makers — a staff of about 140 people — released nearly 10 nanodegrees in 2018. Today, as a result of cuts, only 40 content creators remain. That smaller team completed about five nanodegrees in the first quarter of 2019, Thrun said.

Last year, it took between 10 to 12 people, and more than $1 million, to build one nanodegree, Thrun said. “Now it’s less than 10 percent of that.”

The company was able to accomplish this, he said, by changing its whole approach to video with taping, edits and student assessments happening in real time.

Udacity, under Thrun’s direction, has also doubled down on a technical mentorship program that will now match every new student with a mentor. Udacity has hired about 278 mentors who will work between 15 and 20 hours a week on a contract basis. The company is targeting about 349 mentors in all.

Students are also assigned a cohort that is required to meet (virtually) once a week.

Thrun described the new mentor program as the biggest change in service in the entire history of Udacity. “And we literally did this in two weeks,” he said. 

The strategy has met with some resistance. Some employees wanted to test the mentorship program on one cohort, or group of students, and expand from there. Even since these recent changes, some employees have expressed doubts that it will be enough, according to unnamed sources connected to or within the company.

Even Thrun admits that the “fruit remains to be seen,” although he’s confident that they’ve landed on the right approach, and one that will boost student graduation rates and eventually make the company profitable.

“If you give any student a personalized mentor that fights for them, and that’s the language I usually use, then we can bring our graduation rate, which is at about 34 percent to 60 percent or so,” he said. “And for online institutions 34 percent is high. But we have programs in that graduate more than 90 percent of our students.”

Udacity doesn’t share exact numbers on post-graduation hiring rates. But the company did say thousands of Udacity alumni have been hired by companies like Google, AT&T, Nvidia and others in the U.S., Europe, India and China.

In the U.S. and Canada, graduates with new jobs reported an annual salary increase of 38 percent, a Udacity spokesperson shared.

Indeed, Udacity has had some successes despite its many challenges.

Bright spots

Udacity has continued to increase revenue, although at a slower rate than the previous year-over-year time period. Udacity said it generated $90 million in revenue in 2018, a 25 percent year-over-year increase from 2017.

Udacity had informally offered enterprise programs to clients like AT&T. But in September, the company made enterprise a dedicated product and hired a VP of sales to bring in new customers.

Udacity has added 20 new enterprise clients from the banking, insurance, telecom and retail sectors, according to the company. There are now 70 enterprise customers globally that send employees through Udacity programs to gain new skills.

It continues to expand its career services and launched 12 free courses, built in collaboration with Google, with nearly 100,000 enrollments. It has also funded more than 1.1 million new partial and full scholarships to its programs for students across North America, Europe, the Middle East and Asia. About 21% of all Udacity Nanodegree students in the Grow with Google program in Europe have received job offers, according to Google.

The company also has a new initiative in the Middle East, where it teaches almost a million young Arab people how to code, Thrun said, an accomplishment he says is core to his mission.

Udacity isn’t profitable yet on an EBITDA basis, Thrun shared, but the “unit economics per students, and on a gross margin basis, are good.”

Now, it comes down to whether Thrun’s push to become faster, more efficient and nimble, all while investing in student services and its enterprise product, will be enough to right the ship.

“I really believe if you can get to the point that students come to us and we bend over backwards to ensure their success, we will be a company that has a really good chance of lasting for a lifetime,” he said. 

And if it doesn’t work, then we’ll adjust, like any other company. We can always shift.”


Read Full Article

Learning to Generalize from Sparse and Underspecified Rewards




Reinforcement learning (RL) presents a unified and flexible framework for optimizing goal-oriented behavior, and has enabled remarkable success in addressing challenging tasks such as playing video games, continuous control, and robotic learning. The success of RL algorithms in these application domains often hinges on the availability of high-quality and dense reward feedback. However, broadening the applicability of RL algorithms to environments with sparse and underspecified rewards is an ongoing challenge, requiring a learning agent to generalize (i.e., learn the right behavior) from limited feedback. A natural way to investigate the performance of RL algorithms in such problem settings is via language understanding tasks, where an agent is provided with a natural language input and needs to generate a complex response to achieve a goal specified in the input, while only receiving binary success-failure feedback.

For instance, consider a "blind" agent tasked with reaching a goal position in a maze by following a sequence of natural language commands (e.g., "Right, Up, Up, Right"). Given the input text, the agent (green circle) needs to interpret the commands and take actions based on such interpretation to generate an action sequence (a). The agent receives a reward of 1 if it reaches the goal (red star) and 0 otherwise. Because the agent doesn't have access to any visual information, the only way for the agent to solve this task and generalize to novel instructions is by correctly interpreting the instructions.
In this instruction-following task, the action trajectories a1, a2 and a3 reach the goal, but the sequences a2 and a3 do not follow the instructions. This illustrates the issue of underspecified rewards.
In these tasks, the RL agent needs to learn to generalize from sparse (only a few trajectories lead to a non-zero reward) and underspecified (no distinction between purposeful and accidental success) rewards. Importantly, because of underspecified rewards, the agent may receive positive feedback for exploiting spurious patterns in the environment. This can lead to reward hacking, causing unintended and harmful behavior when deployed in real-world systems.

In "Learning to Generalize from Sparse and Underspecified Rewards", we address the issue of underspecified rewards by developing Meta Reward Learning (MeRL), which provides more refined feedback to the agent by optimizing an auxiliary reward function. MeRL is combined with a memory buffer of successful trajectories collected using a novel exploration strategy to learn from sparse rewards. The effectiveness of our approach is demonstrated on semantic parsing, where the goal is to learn a mapping from natural language to logical forms (e.g., mapping questions to SQL programs). In the paper, we investigate the weakly-supervised problem setting, where the goal is to automatically discover logical programs from question-answer pairs, without any form of program supervision. For instance, given the question "Which nation won the most silver medals?" and a relevant Wikipedia table, an agent needs to generate an SQL-like program that results in the correct answer (i.e., "Nigeria").
The proposed approach achieves state-of-the-art results on the WikiTableQuestions and WikiSQL benchmarks, improving upon prior work by 1.2% and 2.4% respectively. MeRL automatically learns the auxiliary reward function without using any expert demonstrations, (e.g., ground-truth programs) making it more widely applicable and distinct from previous reward learning approaches. The diagram below depicts a high level overview of our approach:
Overview of the proposed approach. We employ (1) mode covering exploration to collect a diverse set of successful trajectories in a memory buffer; (2) Meta-learning or Bayesian optimization to learn an auxiliary reward that provides more refined feedback for policy optimization.
Meta Reward Learning (MeRL)
The key insight of MeRL in dealing with underspecified rewards is that spurious trajectories and programs that achieve accidental success are detrimental to the agent's generalization performance. For example, an agent might be able to solve a specific instance of the maze problem above. However, if it learns to perform spurious actions during training, it is likely to fail when provided with unseen instructions. To mitigate this issue, MeRL optimizes a more refined auxiliary reward function, which can differentiate between accidental and purposeful success based on features of action trajectories. The auxiliary reward is optimized by maximizing the trained agent's performance on a hold-out validation set via meta learning.
Schematic illustration of MeRL: The RL agent is trained via the reward signal obtained from the auxiliary reward model while the auxiliary rewards are trained using the generalization error of the agent.
Learning from Sparse Rewards
To learn from sparse rewards, effective exploration is critical to find a set of successful trajectories. Our paper addresses this challenge by utilizing the two directions of Kullback–Leibler (KL) divergence, a measure on how different two probability distributions are. In the example below, we use KL divergence to minimize the difference between a fixed bimodal (shaded purple) and a learned gaussian (shaded green) distribution, which can represent the distribution of the agent's optimal policy and our learned policy respectively. One direction of the KL objective learns a distribution which tries to cover both the modes while the distribution learned by other objective seeks a particular mode (i.e. it prefers one mode over another). Our method exploits the mode covering KL's tendency to focus on multiple peaks to collect a diverse set of successful trajectories and mode seeking KL's implicit preference between trajectories to learn a robust policy.
Left: Optimizing mode covering KL. Right: Optimizing mode seeking KL

Conclusion
Designing reward functions that distinguish between optimal and suboptimal behavior is critical for applying RL to real-world applications. This research takes a small step in the direction of modelling reward functions without any human supervision. In future work, we'd like to tackle the credit-assignment problem in RL from the perspective of automatically learning a dense reward function.

Acknowledgements
This research was done in collaboration with Chen Liang and Dale Schuurmans. We thank Chelsea Finn and Kelvin Guu for their review of the paper.

Opera Touch brings website cookie blocking to iOS


Last fall, Opera introduced Opera Touch for iOS – a solid alternative to Safari on iPhone, optimized for one-handed use. Today, the company is rolling out a notable new feature to this app: cookie blocking. Yes, it can now block those annoying dialogs that ask you to accept the website’s cookies. These are particularly problematic on mobile, where they often entirely interrupt your ability to view the content, as opposed to on many desktop websites where you can (kind of) ignore the pop-up banner that appears at the bottom or the top of the page.

Cookie dialogs have become prevalent across the web as a result of Europe’s GDPR, but many people find them overly intrusive. Today, it takes an extra click to dismiss these prompts, which slows down web browsing – especially for those times you’re on the hunt for a particular piece of information and are visiting several websites in rapid succession.

The cookie blocking feature was first launched in November on Opera’s flagship app for Android, but hadn’t yet made its way to iOS – through any browser app, that is, not just one from Opera. The company says it uses a mix of CSS and JavaScript heuristics in order to block the prompts.

At the time of the launch, Opera noted it had tested the feature with some 15,000 sites.

It’s important to note that the default setting for the cookie blocker on Opera Touch will allow the websites to set cookies.

Here’s how it works. When you enable the feature, it will hide the dialog boxes from appearing, allowing you to read a website without having to first close the prompt. However, when you turn on the Cookie Blocker option, another setting is also switched on: one that says “automatically accept cookie dialogs.”

That means, in practice, when you’re enabling the Cookie Blocker, you’re also enabling cookie acceptance if you don’t take further action.

But Opera says you can disable this checkbox, if you don’t want your browser to give websites your acceptance.

In addition to the new cookie blocking, the browser has a number of other options that make it an interesting alternative to Safari on iOS or Google Chrome.

For example, if offers built-in ad blocking, cryptocurrency mining protection (which prevents malicious sites from using your device’s resources to mine for cryptocurrencies), a way to send web content to your PC through Opera’s “Flow” technology, and – most importantly – a design focused on using the app with just one hand.

Since the app’s launch in April, the company has rolled out 23 new features in total. This include a new dark theme, as well as the addition of a private mode, plus search engine choice which offers 11 options, including Qwant and DuckDuckGo, and other features.

The app is a free download on iOS.


Read Full Article