Y Combinator is launching a startup program in China

U.S. accelerator Y Combinator is expanding to China after it announced the hiring of former Microsoft and Baidu executive Qi Lu who will develop a standalone startup program that runs on Chinese soil. Shanghai-born Lu spent 11 years with Yahoo and eight years with Microsoft before a short spell with Baidu, where he was COO and […]

U.S. accelerator Y Combinator is expanding to China after it announced the hiring of former Microsoft and Baidu executive Qi Lu who will develop a standalone startup program that runs on Chinese soil.

Shanghai-born Lu spent 11 years with Yahoo and eight years with Microsoft before a short spell with Baidu, where he was COO and head of the firm’s AI research division. Now he becomes founding CEO of YC China while he’s also stepping into the role of Head of YC Research. YC will also expand its research team with an office in Seattle, where Lu has plenty of links.

There’s no immediate timeframe for when YC will launch its China program, which represents its first global expansion, but YC President Sam Altman told TechCrunch in an interview that the program will be based in Beijing once it is up and running. Altman said Lu will use his network and YC’s growing presence in China — it ran its first ‘Startup School’ event in Beijing earlier this year — to recruit prospects who will be put into the upcoming winter program in the U.S..

Following that, YC will work to launch the China-based program as soon as possible. It appears that the details are still being sketched out, although Altman did confirm it will run independently but may lean on local partners for help. The YC President he envisages batch programming in the U.S. and China overlapping to a point with visitors, shared mentors and potentially other interaction between the two.

China’s startup scene has grown massively in recent years, numerous reports peg it close to that of the U.S., so it makes sense that YC, as an ‘ecosystem builder,’ wants to in. But Altman believes that the benefits extend beyond YC and will strengthen its network of founders, which spans more than 1,700 startups.

“The number one asset YC has is a very special founder community,” he told TechCrunch. “The opportunity to include a lot more Chinese founders seems super valuable to everyone. Over the next decade, a significant portion of the tech companies started will be from the U.S. or China [so operating a] network across both is a huge deal.”

Altman said he’s also banking on Lu being the man to make YC China happen. He revealed that he’s spent a decade trying to hire Lu, who he described as “one of the most impressive technologists I know.”

Y Combinator President Sam Altman has often spoken of his desire to get into the Chinese market

Entering China as a foreign entity is never easy, and in the venture world it is particularly tricky because China already has an advanced ecosystem of firms with their own networks for founders, particularly in the early-stage space. But Altman is confident that YC’s global reach and roster of founders and mentors appeals to startups in China.

YC has been working to add Chinese startups to its U.S.-based programs for some time. Altman has long been keen on an expansion to China, as he discussed at our Disrupt event last year, and partner Eric Migicovsky — who co-founder Pebble — has been busy developing networks and arranging events like the Beijing one to raise its profile.

That’s seen some progress with more teams from China — and other parts of the world — taking part in YC batches, which have never been more diverse. But YC is still missing out on global talent.

According to its own data, fewer than 10 Chinese companies have passed through its corridors but that list looks like it is missing some names so the number may be higher. Clearly, though, admission are skewed towards the U.S. — the question is whether Qi Lu and creation of YC China can significantly alter that.

Finding the Goldilocks zone for applied AI

To find the right opportunity around which to build an AI business, startups must apply the “Goldilocks” principle to find the sweet spot that is “just right.”

While Elon Musk and Mark Zuckerberg debate the dangers of artificial general intelligence, startups applying AI to more narrowly defined problems such as accelerating the performance of sales teams and improving the operating efficiency of manufacturing lines are building billion-dollar businesses. Narrowly defining a problem, however, is only the first step to finding valuable business applications of AI.

To find the right opportunity around which to build an AI business, startups must apply the “Goldilocks principle” in several different dimensions to find the sweet spot that is “just right” to begin — not too far in one dimension, not too far in another. Here are some ways for aspiring startup founders to thread the needle with their AI strategy, based on what we’ve learned from working with thousands of AI startups.

 “Just right” prediction time horizons

Unlike pre-intelligence software, AI responds to the environment in which they operate; algorithms take in data and return an answer or prediction. Depending on the application, that prediction may describe an outcome in the near term, such as tomorrow’s weather, or an outcome many years in the future, such as whether a patient will develop cancer in 20 years. The time horizon of the algorithm’s prediction is critical to its usefulness and to whether it offers an opportunity to build defensibility.

Algorithms making predictions with long time horizons are difficult to evaluate and improve. For example, an algorithm may use the schedule of a contractor’s previous projects to predict that a particular construction project will fall six months behind schedule and go over budget by 20 percent. Until this new project is completed, the algorithm designer and end user can only tell whether the prediction is directionally correct — that is, whether the project is falling behind or costs are higher.

Even when the final project numbers end up very close to the predicted numbers, it will be difficult to complete the feedback loop and positively reinforce the algorithm. Many factors may influence complex systems like a construction project, making it difficult to A/B test the prediction to tease out the input variables from unknown confounding factors. The more complex the system, the longer it may take the algorithm to complete a reinforcement cycle, and the more difficult it becomes to precisely train the algorithm.

While many enterprise customers are open to piloting AI solutions, startups must be able to validate the algorithm’s performance in order to complete the sale. The most convincing way to validate an algorithm is by using the customer’s real-time data, but this approach may be difficult to achieve during a pilot. If the startup does get access to the customer’s data, the prediction time horizon should be short enough that the algorithm can be validated during the pilot period.

For most of AI history, slow computational speeds have severely limited the scope of applied AI.

Historic data, if it’s available, can serve as a stopgap to train an algorithm and temporarily validate it via backtesting. Training an algorithm making long time horizon predictions on historic data is risky because processes and environments are more likely to have changed the further back you dig into historic records, making historic data sets less descriptive of present-day conditions.

In other cases, while the historic data describing outcomes exists for you to train an algorithm, it may not capture the input variable under consideration. In the construction example, that could mean that you found out that sites using blue safety hats are more likely to complete projects on time, but since that hat color wasn’t previously helpful in managing projects, that information wasn’t recorded in the archival records. This data must be captured from scratch, which further delays your time to market.

Instead of making singular “hero” predictions with long time horizons, AI startups should build multiple algorithms making smaller, simpler predictions with short time horizons. Decomposing an environment into simpler subsystems or processes limits the number of inputs, making them easier to control for confounding factors. The BIM 360 Project IQ Team at Autodesk takes this small prediction approach to areas that contribute to construction project delays. Their models predict safety and score vendor and subcontractor quality/reliability, all of which can be measured while a project is ongoing.

Shorter time horizons make it easier for the algorithm engineer to monitor its change in performance and take action to quickly improve it, instead of being limited to backtesting on historic data. The shorter the time horizon, the shorter the algorithm’s feedback loop will be. As each cycle through the feedback incrementally compounds the algorithm’s performance, shorter feedback loops are better for building defensibility. 

“Just right” actionability window

Most algorithms model dynamic systems and return a prediction for a human to act on. Depending on how quickly the system is changing, the algorithm’s output may not remain valid for very long: the prediction may “decay” before the user can take action. In order to be useful to the end user, the algorithm must be designed to accommodate the limitations of computing and human speed. 

In a typical AI-human workflow, the human feeds input data into the algorithm, the algorithm runs calculations on that input data and returns an output that predicts a certain outcome or recommends a course of action; the human interprets that information to decide on a course of action, then takes action. The time it takes the algorithm to compute an answer and the time it takes for a human to act on the output are the two largest bottlenecks in this workflow. 

For most of AI history, slow computational speeds have severely limited the scope of applied AI. An algorithm’s prediction depends on the input data, and the input data represents a snapshot in time at the moment it was recorded. If the environment described by the data changes faster than the algorithm can compute the input data, by the time the algorithm completes its computations and returns a prediction, the prediction will only describe a moment in the past and will not be actionable. For example, the algorithm behind the music app Shazam may have needed several hours to identify a song after first “hearing” it using the computational power of a Windows 95 computer. 

The rise of cloud computing and the development of hardware specially optimized for AI computations has dramatically broadened the scope of areas where applied AI is actionable and affordable. While macro tech advancements can greatly advance applied AI, the algorithm is not totally held hostage to current limits of computation; reinforcement through training also can improve the algorithm’s response time. The more of the same example an algorithm encounters, the more quickly it can skip computations to arrive at a prediction. Thanks to advances in computation and reinforcement, today Shazam takes less than 15 seconds to identify a song. 

Automating the decision and action also could help users make use of predictions that decay too quickly to wait for humans to respond. Opsani is one such company using AI to make decisions that are too numerous and fast-moving for humans to make effectively. Unlike human DevOps, who can only move so fast to optimize performance based on recommendations from an algorithm, Opsani applies AI to both identify and automatically improve operations of applications and cloud infrastructure so its customers can enjoy dramatically better performance.

Not all applications of AI can be completely automated, however, if the perceived risk is too high for end users to accept, or if regulations mandate that humans must approve the decision. 

“Just right” performance minimums

Just like software startups launch when they have built a minimum viable product (MVP) in order to collect actionable feedback from initial customers, AI startups should launch when they reach the minimum algorithmic performance (MAP) required by early adopters, so that the algorithm can be trained on more diverse and fresh data sets and avoid becoming overfit to a training set.

Most applications don’t require 100 percent accuracy to be valuable. For example, a fraud detection algorithm may only immediately catch five percent of fraud cases within 24 hours of when they occur, but human fraud investigators catch 15 percent of fraud cases after a month of analysis. In this case, the MAP is zero, because the fraud detection algorithm could serve as a first filter in order to reduce the number of cases the human investigators must process. The startup can go to market immediately in order to secure access to the large volume of fraud data used for training their algorithm. Over time, the algorithms’ accuracy will improve and reduce the burden on human investigators, freeing them to focus on the most complex cases.

Startups building algorithms for zero or low MAP applications will be able to launch quickly, but may be continuously looking over their shoulder for copycats, if these copycats appear before the algorithm has reached a high level of performance. 

There’s no one-size-fits-all approach to moving an algorithm from the research lab to the market.

Startups attacking low MAP problems also should watch out for problems that can be solved with near 100 percent accuracy with a very small training set, where the problem being modeled is relatively simple, with few dimensions to track and few possible variations in outcome.

AI-powered contract processing is a good example of an application where the algorithm’s performance plateaus quickly. There are thousands of contract types, but most of them share key fields: the parties involved, the items of value being exchanged, time frame, etc. Specific document types like mortgage applications or rental agreements are highly standardized in order to comply with regulation. Across multiple startups, we have seen algorithms that automatically process these documents needing only a few hundred examples to train to an acceptable degree of accuracy before additional examples do little to improve the algorithm, making it easy for new entrants to match incumbents and earlier entrants in performance.

AIs built for applications where human labor is inexpensive and able to easily achieve high accuracy may need to reach a higher MAP before they can find an early adopter. Tasks requiring fine motor skills, for example, have yet to be taken over by robots because human performance sets a very high MAP to overcome. When picking up an object, the AIs powering the robotic hand must gauge an object’s stiffness and weight with a high degree of accuracy, otherwise the hand will damage the object being handled. Humans can very accurately gauge these dimensions with almost no training. Startups attacking high MAP problems must invest more time and capital into acquiring enough data to reach MAP and launch. 

Threading the needle

Narrow AI can demonstrate impressive gains in a wide range of applications — in the research lab. Building a business around a narrow AI application, on the other hand, requires a new playbook. This process is heavily dependent on the specific use case on all dimensions, and the performance of the algorithm is merely one starting point. There’s no one-size-fits-all approach to moving an algorithm from the research lab to the market, but we hope these ideas will provide a useful blueprint for you to begin.

Nvidia’s new Turing architecture is all about real-time ray tracing and AI

In recent days, word about Nvidia’s new Turing architecture started leaking out of the Santa Clara-based company’s headquarters. So it didn’t come as a major surprise that the company today announced during its Siggraph keynote the launch of this new architecture and three new pro-oriented workstation graphics cards in its Quadro family. Nvidia describes the […]

In recent days, word about Nvidia’s new Turing architecture started leaking out of the Santa Clara-based company’s headquarters. So it didn’t come as a major surprise that the company today announced during its Siggraph keynote the launch of this new architecture and three new pro-oriented workstation graphics cards in its Quadro family.

Nvidia describes the new Turing architecture as “the greatest leap since the invention of the CUDA GPU in 2006.” That’s a high bar to clear, but there may be a kernel of truth here. These new Quadro RTx chips are the first to feature the company’s new RT Cores. “RT” here stands for ray tracing, a rendering method that basically traces the path of light as it interacts with the objects in a scene. This technique has been around for a very long time (remember POV-Ray on the Amiga?). Traditionally, though, it was always very computationally intensive, though the results tend to look far more realistic. In recent years, ray tracing got a new boost thanks to faster GPUs and support from the likes of Microsoft, which recently added ray tracing support to DirectX.

“Hybrid rendering will change the industry, opening up amazing possibilities that enhance our lives with more beautiful designs, richer entertainment and more interactive experiences,” said Nvidia CEO Jensen Huang. “The arrival of real-time ray tracing is the Holy Grail of our industry.”

The new RT cores can accelerate ray tracing by up to 25 times compared to Nvidia’s Pascal architecture, and Nvidia claims 10 GigaRays a second for the maximum performance.

Unsurprisingly, the three new Turing-based Quadro GPUs will also feature the company’s AI-centric Tensor Cores, as well as 4,608 CUDA cores that can deliver up to 16 trillion floating point operations in parallel with 16 trillion integer operations per second. The chips feature GDDR6 memory to expedite things, and support Nvidia’s NVLink technology to scale up memory capacity to up to 96GB and 100GB/s of bandwidth.

The AI part here is more important than it may seem at first. With NGX, Nvidia today also launched a new platform that aims to bring AI into the graphics pipelines. “NGX technology brings capabilities such as taking a standard camera feed and creating super slow motion like you’d get from a $100,000+ specialized camera,” the company explains, and also notes that filmmakers could use this technology to easily remove wires from photographs or replace missing pixels with the right background.

On the software side, Nvidia also today announced that it is open sourcing its Material Definition Language (MDL).

Companies ranging from Adobe (for Dimension CC) to Pixar, Siemens, Black Magic, Weta Digital, Epic Games and Autodesk have already signed up to support the new Turing architecture.

All of this power comes at a price, of course. The new Quadro RTX line starts at $2,300 for a 16GB version, while stepping up to 24GB will set you back $6,300. Double that memory to 48GB and Nvidia expects that you’ll pay about $10,000 for this high-end card.

Observe.AI raises $8M to use artificial intelligence to improve call centers

Being stuck on the phone with call centers is painful. We all know this. Observe.AI is one company that wants to make the experience more bearable, and it’s raised $8 million to develop an artificial intelligence system that it believes will do just that. The funding round was led by Nexus Venture Partners, with participation from MGV, Liquid 2 […]

Being stuck on the phone with call centers is painful. We all know this. Observe.AI is one company that wants to make the experience more bearable, and it’s raised $8 million to develop an artificial intelligence system that it believes will do just that.

The funding round was led by Nexus Venture Partners, with participation from MGV, Liquid 2 Ventures and Hack VC. Existing investors Emergent Ventures and Y Combinator also took part — Observe.AI was part of the YC’s winter 2018 batch.

The India-U.S. startup was founded last year with the goal of solving a very personal problem for founders Swapnil Jain (CEO), Akash Singh (CTO) and Sharath Keshava (CRO): making call centers better. But, unlike most AI products that offer the potential to fully replace human workforces, Observe.AI is setting out to help the humble customer service agent.

The company’s first product is an AI that assists call center workers by automating a range of tasks, from auto-completing forms for customers to guiding them on next steps in-call and helping find information quickly. Jain told TechCrunch in an interview that the product was developed following months of consultation with call center companies and their staff, both senior and junior. That included a stint in Manila, one of the world’s capitals for offshoring customer services and a city well known to Keshava, who helped healthcare startup Practo launch its business in the Philippines’ capital.

That effort to know call center operates directly has also shaped how Observe.AI is pitching its services. Rather than going to companies, it is tapping the root of the tree by offering its services to the call centers who manage customer support for well-known businesses behind the curtain. Uber, for example, is one of many to use Philippines-based support centers, but the Observe.AI thesis is that going directly to the source is easier than navigating large companies for business.

One such partner is Concentrix, one of the world’s largest customer support providers with over 100,000 staff and offices dotted around the globe, while the startup said it has tapped Philippines telco PLDT for infrastructure.

In addition to helping understand the problems and generating business, working directly with these companies also gives Observe.AI access to and use of data, which is essential for developing any AI and natural language processing-based systems.

Beyond improving its customer service assistant — which Jain likens to an ‘Alexa for call centers’ — Observe.AI is working to develop a virtual assistant of its own that can handle the more basic and repetitive calls from customers to help free up agents for callers who need a human on the other end of the line.

“We aim to eventually automate a large part of the call center experience,” Jain explained in an interview. “A good set [of customer calls] are complex but a large set can be fairly automated as they are simple in nature.”

The startup is aiming to introduce ‘voicebots’ before March 2020, with a beta launch targeted at the end of 2019.

“The kind of company that will disrupt call centers will come from the east — we truly understand the call center life,” Jain told TechCrunch.

He explained that, while Silicon Valley is a hotbed for tech development, understanding the problems that need to be solved requires spending time in markets like India and the Philippines.

“That knowledge is super, super valuable… someone in the U.S. can’t even think about it,” he added.

That said, Observe.AI is headquartered in the U.S., in Santa Clara. That’s where Keshava, the company CRO, is based with a growing team that is dedicated pre- and post-sales and to building relationships with major software platforms used by call center companies. The idea with the latter is that they can provide an avenue into new business by working with Observe.AI to add AI smarts to their product.

In one such example, Talkdesk, a U.S. startup that offers cloud-based contact center services, has added Observe.AI’s services to what it offers to its customers. Talkdesk CEO Tiago Paiva called the addition “a huge opportunity for call center efficiency and improving the caller experience.”

The startup’s India-based team is Bangalore and it handles technology, which includes the machine learning component. Total headcount is 16 people right now but the founding team expects that will at least double before the end of this year.

New Uber feature uses machine learning to sort business and personal rides

Uber announced a new program today called Profile Recommendations that takes advantage of machine intelligence to reduce user error when switching between personal and business accounts. It’s not unusual for a person to have both types of accounts. When you’re out and about, it’s easy to forget to switch between them when appropriate. Uber wants to […]

Uber announced a new program today called Profile Recommendations that takes advantage of machine intelligence to reduce user error when switching between personal and business accounts.

It’s not unusual for a person to have both types of accounts. When you’re out and about, it’s easy to forget to switch between them when appropriate. Uber wants to help by recommending the correct one.

“Using machine learning, Uber can predict which profile and corresponding payment method an employee should be using, and make the appropriate recommendation,” Ronnie Gurion, GM and Global Head of Uber for Business wrote in a blog post announcing the new feature.

Uber has been analyzing a dizzying amount of trip data for so long, it can now (mostly) understand the purpose of a given trip based on the details of your request. While it’s certainly not perfect because it’s not always obvious what the purpose is, Uber believes it can determine the correct intention 80 percent of the time. For that remaining 20 percent, when it doesn’t get it right, Uber is hoping to simplify corrections too.

Photo: Uber

Business users can now also assign trip reviewers — managers or other employees who understand the employee’s usage patterns, and can flag questionable rides. Instead of starting an email thread or complicated bureaucratic process to resolve an issue, the employee can now see these flagged rides and resolve them right in the app. “This new feature not only saves the employee’s and administrator’s time, but it also cuts down on delays associated with closing out reports,” Gurion wrote in the blog post announcement.

Uber also announced that it’s supporting a slew of new expense reporting software to simplify integration with these systems. They currently have integrations with Certify, Chrome River, Concur and Expensify. They will be adding support for Expensya, Happay, Rydoo, Zeno by Serko and Zoho Expense starting in September.

All of this should help business account holders deal with Uber expenses more efficiently, while integrating with many of the leading expense programs to move data smoothly from Uber to a company’s regular record-keeping systems.

Openbook is the latest dream of a digital life beyond Facebook

As tech’s social giants wrestle with antisocial demons that appear to be both an emergent property of their platform power, and a consequence of specific leadership and values failures (evident as they publicly fail to enforce even the standards they claim to have), there are still people dreaming of a better way. Of social networking beyond outrage-fuelled […]

As tech’s social giants wrestle with antisocial demons that appear to be both an emergent property of their platform power, and a consequence of specific leadership and values failures (evident as they publicly fail to enforce even the standards they claim to have), there are still people dreaming of a better way. Of social networking beyond outrage-fuelled adtech giants like Facebook and Twitter.

There have been many such attempts to build a ‘better’ social network of course. Most have ended in the deadpool. A few are still around with varying degrees of success/usage (Snapchat, Ello and Mastodon are three that spring to mine). None has usurped Zuckerberg’s throne of course.

This is principally because Facebook acquired Instagram and WhatsApp. It has also bought and closed down smaller potential future rivals (tbh). So by hogging network power, and the resources that flow from that, Facebook the company continues to dominate the social space. But that doesn’t stop people imagining something better — a platform that could win friends and influence the mainstream by being better ethically and in terms of functionality.

And so meet the latest dreamer with a double-sided social mission: Openbook.

The idea (currently it’s just that; a small self-funded team; a manifesto; a prototype; a nearly spent Kickstarter campaign; and, well, a lot of hopeful ambition) is to build an open source platform that rethinks social networking to make it friendly and customizable, rather than sticky and creepy.

Their vision to protect privacy as a for-profit platform involves a business model that’s based on honest fees — and an on-platform digital currency — rather than ever watchful ads and trackers.

There’s nothing exactly new in any of their core ideas. But in the face of massive and flagrant data misuse by platform giants these are ideas that seem to sound increasingly like sense. So the element of timing is perhaps the most notable thing here — with Facebook facing greater scrutiny than ever before, and even taking some hits to user growth and to its perceived valuation as a result of ongoing failures of leadership and a management philosophy that’s been attacked by at least one of its outgoing senior execs as manipulative and ethically out of touch.

The Openbook vision of a better way belongs to Joel Hernández who has been dreaming for a couple of years, brainstorming ideas on the side of other projects, and gathering similarly minded people around him to collectively come up with an alternative social network manifesto — whose primary pledge is a commitment to be honest.

“And then the data scandals started happening and every time they would, they would give me hope. Hope that existing social networks were not a given and immutable thing, that they could be changed, improved, replaced,” he tells TechCrunch.

Rather ironically Hernández says it was overhearing the lunchtime conversation of a group of people sitting near him — complaining about a laundry list of social networking ills; “creepy ads, being spammed with messages and notifications all the time, constantly seeing the same kind of content in their newsfeed” — that gave him the final push to pick up the paper manifesto and have a go at actually building (or, well, trying to fund building… ) an alternative platform. 

At the time of writing Openbook’s Kickstarter crowdfunding campaign has a handful of days to go and is only around a third of the way to reaching its (modest) target of $115k, with just over 1,000 backers chipping in. So the funding challenge is looking tough.

The team behind Openbook includes crypto(graphy) royalty, Phil Zimmermann — aka the father of PGP — who is on board as an advisor initially but billed as its “chief cryptographer”, as that’s what he’d be building for the platform if/when the time came. 

Hernández worked with Zimmermann at the Dutch telecom KPN building security and privacy tools for internal usage — so called him up and invited him for a coffee to get his thoughts on the idea.

“As soon as I opened the website with the name Openbook, his face lit up like I had never seen before,” says Hernández. “You see, he wanted to use Facebook. He lives far away from his family and facebook was the way to stay in the loop with his family. But using it would also mean giving away his privacy and therefore accepting defeat on his life-long fight for it, so he never did. He was thrilled at the possibility of an actual alternative.”

On the Kickstarter page there’s a video of Zimmermann explaining the ills of the current landscape of for-profit social platforms, as he views it. “If you go back a century, Coca Cola had cocaine in it and we were giving it to children,” he says here. “It’s crazy what we were doing a century ago. I think there will come a time, some years in the future, when we’re going to look back on social networks today, and what we were doing to ourselves, the harm we were doing to ourselves with social networks.”

“We need an alternative to the social network work revenue model that we have today,” he adds. “The problem with having these deep machine learning neural nets that are monitoring our behaviour and pulling us into deeper and deeper engagement is they already seem to know that nothing drives engagement as much as outrage.

“And this outrage deepens the political divides in our culture, it creates attack vectors against democratic institutions, it undermines our elections, it makes people angry at each other and provides opportunities to divide us. And that’s in addition to the destruction of our privacy by revenue models that are all about exploiting our personal information. So we need some alternative to this.”

Hernández actually pinged TechCrunch’s tips line back in April — soon after the Cambridge Analytica Facebook scandal went global — saying “we’re building the first ever privacy and security first, open-source, social network”.

We’ve heard plenty of similar pitches before, of course. Yet Facebook has continued to harvest global eyeballs by the billions. And even now, after a string of massive data and ethics scandals, it’s all but impossible to imagine users leaving the site en masse. Such is the powerful lock-in of The Social Network effect.

Regulation could present a greater threat to Facebook, though others argue more rules will simply cement its current dominance.

Openbook’s challenger idea is to apply product innovation to try to unstick Zuckerberg. Aka “building functionality that could stand for itself”, as Hernández puts it.

“We openly recognise that privacy will never be enough to get any significant user share from existing social networks,” he says. “That’s why we want to create a more customisable, fun and overall social experience. We won’t follow the footsteps of existing social networks.”

Data portability is an important ingredient to even being able to dream this dream — getting people to switch from a dominant network is hard enough without having to ask them to leave all their stuff behind as well as their friends. Which means that “making the transition process as smooth as possible” is another project focus.

Hernández says they’re building data importers that can parse the archive users are able to request from their existing social networks — to “tell you what’s in there and allow you to select what you want to import into Openbook”.

These sorts of efforts are aided by updated regulations in Europe — which bolster portability requirements on controllers of personal data. “I wouldn’t say it made the project possible but… it provided us a with a unique opportunity no other initiative had before,” says Hernández of the EU’s GDPR.

“Whether it will play a significant role in the mass adoption of the network, we can’t tell for sure but it’s simply an opportunity too good to ignore.”

On the product front, he says they have lots of ideas — reeling off a list that includes the likes of “a topic-roulette for chats, embracing Internet challenges as another kind of content, widgets, profile avatars, AR chatrooms…” for starters.

“Some of these might sound silly but the idea is to break the status quo when it comes to the definition of what a social network can do,” he adds.

Asked why he believes other efforts to build ‘ethical’ alternatives to Facebook have failed he argues it’s usually because they’ve focused on technology rather than product.

“This is still the most predominant [reason for failure],” he suggests. “A project comes up offering a radical new way to do social networking behind the scenes. They focus all their efforts in building the brand new tech needed to do the very basic things a social network can already do. Next thing you know, years have passed. They’re still thousands of miles away from anything similar to the functionality of existing social networks and their core supporters have moved into yet another initiative making the same promises. And the cycle goes on.”

He also reckons disruptive efforts have fizzled out because they were too tightly focused on being just a solution to an existing platform problem and nothing more.

So, in other words, people were trying to build an ‘anti-Facebook’, rather than a distinctly interesting service in its own right. (The latter innovation, you could argue, is how Snap managed to carve out a space for itself in spite of Facebook sitting alongside it — even as Facebook has since sought to crush Snap’s creative market opportunity by cloning its products.)

“This one applies not only to social network initiatives but privacy-friendly products too,” argues Hernández. “The problem with that approach is that the problems they solve or claim to solve are most of the time not mainstream. Such as the lack of privacy.

“While these products might do okay with the people that understand the problems, at the end of the day that’s a very tiny percentage of the market. The solution these products often present to this issue is educating the population about the problems. This process takes too long. And in topics like privacy and security, it’s not easy to educate people. They are topics that require a knowledge level beyond the one required to use the technology and are hard to explain with examples without entering into the conspiracy theorist spectrum.”

So the Openbook team’s philosophy is to shake things up by getting people excited for alternative social networking features and opportunities, with merely the added benefit of not being hostile to privacy nor algorithmically chain-linked to stoking fires of human outrage.

The reliance on digital currency for the business model does present another challenge, though, as getting people to buy into this could be tricky. After all payments equal friction.

To begin with, Hernández says the digital currency component of the platform would be used to let users list secondhand items for sale. Down the line, the vision extends to being able to support a community of creators getting a sustainable income — thanks to the same baked in coin mechanism enabling other users to pay to access content or just appreciate it (via a tip).

So, the idea is, that creators on Openbook would be able to benefit from the social network effect via direct financial payments derived from the platform (instead of merely ad-based payments, such as are available to YouTube creators) — albeit, that’s assuming reaching the necessary critical usage mass. Which of course is the really, really tough bit.

“Lower cuts than any existing solution, great content creation tools, great administration and overview panels, fine-grained control over the view-ability of their content and more possibilities for making a stable and predictable income such as creating extra rewards for people that accept to donate for a fixed period of time such as five months instead of a month to month basis,” says Hernández, listing some of the ideas they have to stand out from existing creator platforms.

“Once we have such a platform and people start using tips for this purpose (which is not such a strange use of a digital token), we will start expanding on its capabilities,” he adds. (He’s also written the requisite Medium article discussing some other potential use cases for the digital currency portion of the plan.)

At this nascent prototype and still-not-actually-funded stage they haven’t made any firm technical decisions on this front either. And also don’t want to end up accidentally getting into bed with an unethical tech.

“Digital currency wise, we’re really concerned about the environmental impact and scalability of the blockchain,” he says — which could risk Openbook contradicting stated green aims in its manifesto and looking hypocritical, given its plan is to plough 30% of its revenues into ‘give-back’ projects, such as environmental and sustainability efforts and also education.

“We want a decentralised currency but we don’t want to rush into decisions without some in-depth research. Currently, we’re going through IOTA’s whitepapers,” he adds.

They do also believe in decentralizing the platform — or at least parts of it — though that would not be their first focus on account of the strategic decision to prioritize product. So they’re not going to win fans from the (other) crypto community. Though that’s hardly a big deal given their target user-base is far more mainstream.

“Initially it will be built on a centralised manner. This will allow us to focus in innovating in regards to the user experience and functionality product rather than coming up with a brand new behind the scenes technology,” he says. “In the future, we’re looking into decentralisation from very specific angles and for different things. Application wise, resiliency and data ownership.”

“A project we’re keeping an eye on and that shares some of our vision on this is Tim Berners Lee’s MIT Solid project. It’s all about decoupling applications from the data they use,” he adds.

So that’s the dream. And the dream sounds good and right. The problem is finding enough funding and wider support — call it ‘belief equity’ — in a market so denuded of competitive possibility as a result of monopolistic platform power that few can even dream an alternative digital reality is possible.

In early April, Hernández posted a link to a basic website with details of Openbook to a few online privacy and tech communities asking for feedback. The response was predictably discouraging. “Some 90% of the replies were a mix between critiques and plain discouraging responses such as “keep dreaming”, “it will never happen”, “don’t you have anything better to do”,” he says.

(Asked this April by US lawmakers whether he thinks he has a monopoly, Zuckerberg paused and then quipped: “It certainly doesn’t feel like that to me!”)

Still, Hernández stuck with it, working on a prototype and launching the Kickstarter. He’s got that far — and wants to build so much more — but getting enough people to believe that a better, fairer social network is even possible might be the biggest challenge of all. 

For now, though, Hernández doesn’t want to stop dreaming.

“We are committed to make Openbook happen,” he says. “Our back-up plan involves grants and impact investment capital. Nothing will be as good as getting our first version through Kickstarter though. Kickstarter funding translates to absolute freedom for innovation, no strings attached.”

You can check out the Openbook crowdfunding pitch here.

NASA’s Parker Solar Probe launches tonight to ‘touch the sun’

NASA’s ambitious mission to go closer to the Sun than ever before is set to launch in the small hours between Friday and Saturday — at 3:33 AM Eastern from Kennedy Space Center in Florida, to be precise. The Parker Solar Probe, after a handful of gravity assists and preliminary orbits, will enter a stable orbit around the enormous nuclear fireball that gives us all life and sample its radiation from less than 4 million miles away. Believe me, you don’t want to get much closer than that.

NASA’s ambitious mission to go closer to the Sun than ever before is set to launch in the small hours between Friday and Saturday — at 3:33 AM Eastern from Kennedy Space Center in Florida, to be precise. The Parker Solar Probe, after a handful of gravity assists and preliminary orbits, will enter a stable orbit around the enormous nuclear fireball that gives us all life and sample its radiation from less than 4 million miles away. Believe me, you don’t want to get much closer than that.

If you’re up late tonight (technically tomorrow morning), you can watch the launch live on NASA’s stream.

This is the first mission named after a living researcher, in this case Eugene Parker, who in the ’50s made a number of proposals and theories about the way that stars give off energy. He’s the guy who gave us solar wind, and his research was hugely influential in the study of the sun and other stars — but it’s only now that some of his hypotheses can be tested directly. (Parker himself visited the craft during its construction, and will be at the launch. No doubt he is immensely proud and excited about this whole situation.)

“Directly” means going as close to the sun as technology allows — which leads us to the PSP’s first major innovation: its heat shield, or thermal protection system.

There’s one good thing to be said for the heat near the sun: it’s a dry heat. Because there’s no water vapor or gases in space to heat up, find some shade and you’ll be quite comfortable. So the probe is essentially carrying the most heavy-duty parasol ever created.

It’s a sort of carbon sandwich, with superheated carbon composite on the outside and a carbon foam core. All together it’s less than a foot thick, but it reduces the temperature the probe’s instruments are subjected to from 2,500 degrees Fahrenheit to 85 — actually cooler than it is in much of the U.S. right now.

Go on – it’s quite cool.

The car-sized Parker will orbit the sun and constantly rotate itself so the heat shield is facing inward and blocking the brunt of the solar radiation. The instruments mostly sit behind it in a big insulated bundle.

And such instruments! There are three major experiments or instrument sets on the probe.

WISPR (Wide-Field Imager for Parker Solar Probe) is a pair of wide-field telescopes that will watch and image the structure of the corona and solar wind. This is the kind of observation we’ve made before — but never from up close. We generally are seeing these phenomena from the neighborhood of the Earth, nearly 100 million miles away. You can imagine that cutting out 90 million miles of cosmic dust, interfering radiation and other nuisances will produce an amazingly clear picture.

SWEAP (Solar Wind Electrons Alphas and Protons investigation) looks out to the side of the craft to watch the flows of electrons as they are affected by solar wind and other factors. And on the front is the Solar Probe Cup (I suspect this is a reference to the Ray Bradbury story, “Golden Apples of the Sun”), which is exposed to the full strength of the sun’s radiation; a tiny opening allows charged particles in, and by tracking how they pass through a series of charged windows, they can sort them by type and energy.

FIELDS is another that gets the full heat of the sun. Its antennas are the ones sticking out from the sides — they need to in order to directly sample the electric field surrounding the craft. A set of “fluxgate magnetometers,” clearly a made-up name, measure the magnetic field at an incredibly high rate: two million samples per second.

They’re all powered by solar panels, which seems obvious, but actually it’s a difficult proposition to keep the panels from overloading that close to the sun. They hide behind the shield and just peek out at an oblique angle, so only a fraction of the radiation hits them.

Even then, they’ll get so hot that the team needed to implement the first-ever active water cooling system on a spacecraft. Water is pumped through the cells and back behind the shield, where it is cooled by, well, space.

The probe’s mission profile is a complicated one. After escaping the clutches of the Earth, it will swing by Venus, not to get a gravity boost, but “almost like doing a little handbrake turn,” as one official described it. It slows it down and sends it closer to the sun — and it’ll do that seven more times, each time bringing it closer and closer to the sun’s surface, ultimately arriving in a stable orbit 3.83 million miles above the surface — that’s 95 percent of the way from the Earth to the sun.

On the way it will hit a top speed of 430,000 miles per hour, which will make it the fastest spacecraft ever launched.

Parker will make 24 total passes through the corona, and during these times communication with Earth may be interrupted or impractical. If a solar cell is overheating, do you want to wait 20 minutes for a decision from NASA on whether to pull it back? No. This close to the sun even a slight miscalculation results in the reduction of the probe to a cinder, so the team has imbued it with more than the usual autonomy.

It’s covered in sensors in addition to its instruments, and an onboard AI will be empowered to make decisions to rectify anomalies. That sounds worryingly like a HAL 9000 situation, but there are no humans on board to kill, so it’s probably okay.

The mission is scheduled to last seven years, after which time the fuel used to correct the craft’s orbit and orientation is expected to run out. At that point it will continue as long as it can before drift causes it to break apart and, one rather hopes, become part of the sun’s corona itself.

The Parker Solar Probe is scheduled for launch early Saturday morning, and we’ll update this post when it takes off successfully or, as is possible, is delayed until a later date in the launch window.

At Disrupt, Hans Tung and Yi Wang will talk about the startup road winding from China to the U.S.

Few investors have as deep a knowledge of the U.S. and Chinese markets as Hans Tung. For over a decade the prodigious investor (now with GGV Capital) has been racking up the miles on flights between San Francisco, Shanghai, Los Angeles, Beijing, and New York in search of startups that can span the Pacific divide […]

Few investors have as deep a knowledge of the U.S. and Chinese markets as Hans Tung.

For over a decade the prodigious investor (now with GGV Capital) has been racking up the miles on flights between San Francisco, Shanghai, Los Angeles, Beijing, and New York in search of startups that can span the Pacific divide as readily as he does.

Over time, that’s led to a portfolio that includes Sino-American sweetheart deals in companies like the multi-billion dollar retailer, Wish; the recently acquired social media sensation musical.ly; and the Shanghai social and recommendation service Xiaohongshu; along with U.S.-centric investments like OfferUp and Poshmark.

On stage at Disrupt, Tung will be joined by Yi Wang, the founder of the artificial intelligence-powered education Chinese education dynamo, LingoChamp (Liulishuo), to discuss the technologies and techniques that continue to power a cross-border technology revolution even in tumultuous times.

There could be no better pairing to lead us along the path that winds from the glass and steel campuses of Silicon Valley to the glass and steel office towers of Beijing’s technology parks.

Join us to hear how artificial intelligence is drawing investment dollars on both sides of the Pacific, and how companies are bridging the political divide with compelling new technologies.

It’s sure to be one helluva conversation.

The full agenda is here. You can purchase tickets here.

Siri is now trained to recognize your local, weirdly named small businesses

Getting directions to the nearest Starbucks or Target is a task Apple’s virtual assistant can handle with ease. But what about local businesses with names that Siri has never heard, and might mistake for another phrase or the user misspeaking? To handle these, Apple has created libraries of hyper-local place names so Siri never hears […]

Getting directions to the nearest Starbucks or Target is a task Apple’s virtual assistant can handle with ease. But what about local businesses with names that Siri has never heard, and might mistake for another phrase or the user misspeaking? To handle these, Apple has created libraries of hyper-local place names so Siri never hears “Godfather’s Pizza” as “got father’s piece.”

Speech recognition systems have to be trained on large bodies of data, but while that makes them highly capable when it comes to parsing sentences and recognizing phrases, it doesn’t always teach them the kind of vocabulary that you and your friends use all the time.

When I tell a friend, “let’s go to St John’s for a drink,” they know I don’t mean some cathedral in the midwest but the bar up the street. But Siri doesn’t really have any way of knowing that — in fact, unless the system knows that “Saint John’s” is a phrase in the first place, it might think I’m saying something else entirely. It’s different when you type it into a box — it can just match strings — but when you say it, Siri has to make her best guess at what you said.

But if Siri knew that in the Seattle area, when someone says something that sounds like St John’s, they probably mean the bar, then she can respond more quickly and accurately, without having to think hard or have you select from a list of likely saints. And that’s just what Apple’s latest research does. It’s out now in English, and other languages are likely only a matter of time.

To do this, Apple’s voice recognition team pulled local search results from Apple Maps, sorting out the “places of interest” — you (or an algorithm) can spot these, because people refer to them in certain ways, like “where is the nearest…” and “directions to…” and that sort of thing.

Obviously the sets of these POIs, once you remove national chains like Taco Bell, will represent the unique places that people in a region search for. Burger-seekers here in Seattle will ask about the nearest Dick’s Drive-in, for example (though we already know where they are), while those in L.A. will of course be looking for In-N-Out. But someone in Pittsburgh likely is never looking for either.

Apple sorted these into 170 distinct areas: 169 “combined statistical areas” as defined by the U.S. Census Bureau, which are small enough to have local preferences but not so small that you end up with thousands of them. The special place names for each of these were trained not into the main language model (LM) used by Siri, but into tiny adjunct models (called Geo-LMs) that can be tagged in if the user is looking for a POI using those location-indicating phrases from above.

So when you ask “who is Machiavelli,” you get the normal answer. But when you ask “where is Machiavelli’s,” that prompts the system to query the local Geo-LM (your location is known, of course) and check whether Machiavelli’s is on the list of local POIs (it should be, because the food is great there). Now Siri knows to respond with directions to the restaurant and not to the actual castle where Machiavelli was imprisoned.

Doing this cut the error rate by huge amount – from as much as 25-30 percent to 10-15. That means getting the right result 8 or 9 out of 10 times rather than 2 out of 3; a qualitative improvement that could prevent people from abandoning Siri queries in frustration when it repeatedly fails to understand what they want.

What’s great about this approach is that it’s relatively simple (if not trivial) to expand to other languages and domains. There’s no reason it wouldn’t work for Spanish or Korean, as long as there’s enough data to build it on. And for that matter, why shouldn’t Siri have a special vocabulary set for people in a certain jargon-heavy industry, to reduce spelling errors in notes?

This improved capability is already out, so you should be able to test it out now — or maybe you have been for the last few weeks and didn’t even know it.

AI training and social network content moderation services bring TaskUs a $250 million windfall

TaskUs, the business process outsourcing service that moderates content, annotates information and handles back office customer support for some of the world’s largest tech companies, has raised $250 million in an investment from funds managed by the New York-based private equity giant, Blackstone Group. It’s been ten years since TaskUs was founded with a $20,000 investment […]

TaskUs, the business process outsourcing service that moderates content, annotates information and handles back office customer support for some of the world’s largest tech companies, has raised $250 million in an investment from funds managed by the New York-based private equity giant, Blackstone Group.

It’s been ten years since TaskUs was founded with a $20,000 investment from its two co-founders, and the new deal, which values the decade-old company at $500 million before the money even comes in, is proof of how much has changed for the service in the years since it was founded.

The Santa Monica-based company, which began as a browser-based virtual assistant company — “You send us a task and we get the task done,” recalled TaskUs chief executive Bryce Maddock — is now one of the main providers in the growing field of content moderation for social networks and content annotation for training the algorithms that power artificial intelligence services around the world.

“What I can tell you is we do content moderation for almost every major social network and it’s the fastest growing part of our business today,” Maddock said.

From a network of offices spanning the globe from Mexico to Taiwan and the Philippines to the U.S., the thirty two year-old co-founders Maddock and Jaspar Weir have created a business that’s largest growth stems from snuffing out the distribution of snuff films; child pornography; inappropriate political content and the trails of human trafficking from the user and advertiser generated content on some of the world’s largest social networks.

(For a glimpse into how horrific that process can be, take a look at this article from Wiredwhich looked at content moderation for the anonymous messaging service, Whisper.)

Maddock estimates that while the vast majority of the business was outsourcing business process services in the company’s early days (whether that was transcribing voice mails to texts for the messaging service PhoneTag, or providing customer service and support for companies like HotelTonight) now about 40% of the business comes from content moderation.

Image courtesy of Getty Images

Indeed, it was the growth in new technology services that attracted Blackstone to the business, according to Amit Dixit, Senior Managing Director at Blackstone.

“The growth in ride sharing, social media, online food delivery, e-commerce and autonomous driving is creating an enormous need for enabling business services,” said Dixit in a statement. “TaskUs has established a leadership position in this domain with its base of marquee customers, unique culture, and relentless focus on customer delivery.”

While the back office business processing services remain the majority of the company’s revenue, Maddock knows that the future belongs to an increasing automation of the company’s core services. That’s why part of the money is going to be invested in a new technology integration and consulting business that advises tech companies on which new automation tools to deploy, along with shoring up the company’s position as perhaps the best employer to work for in the world of content moderation and algorithm training services.

It’s been a long five year journey to get to the place it’s in now, with glowing reviews from employees on Glassdoor and social networks like Facebook, Maddock said. The company pays well above minimum wage in the market it operates in (Maddock estimates at least a 50% premium); and provides a generous package of benefits for what Maddock calls the “frontline” teammates. That includes perks like educational scholarships for one child of employees that have been with the company longer than one year; healthcare plans for the employee and three beneficiaries in the Philippines; and 120 days of maternity leave.

And, as content moderation is becoming more automated, the TaskUs employees are spending less time in the human cesspool that attempts to flood social networks every day.

“Increasingly the work that we’re doing is more nuanced. Does this advertisement have political intent. That type of work is far more engaging and could be seen to be a little bit less taxing,” Maddock said.

But he doesn’t deny that the bulk of the hard work his employees are tasked with is identifying and filtering the excremental trash that people would post online.

“I do think that the work is absolutely necessary. The alternative is that everybody has to look at this stuff. it has to be done in a way thats thoughtful and puts the interests of the people who are on the frontlines at the forefront of that effort,” says Maddock. “There have been multiple people who have been involved in sex trafficking, human trafficking and pedophilia that have been arrested directly because of the work that TaskUs is doing. And the consequence of someone not doing that is a far far worse world.”

Maddock also said that TaskUs now shields its employees from having to perform content moderation for an entire shift. “What we have tried to do universally is that there is a subject matter rotation so that you are not just sitting and doing that work all day.”

And the company’s executive knows how taxing the work can be because he said he does it himself. “I try to spend a day a quarter doing the work of our frontline teammates. I spend half my time in our offices,” Maddock said.

Now, with the new investment, TaskUs is looking to expand into additional markets in the UK, Europe, India, and Latin America, Maddock said.

“So far all we’ve been doing is hiring as fast as we possibly can,” said Maddock. “At some point in the future, there’s going to be a point when companies like ours will see the effects of automation,” he added, but that’s why the company is investing in the consulting business… so it can stay ahead of the trends in automation.

Even with the threat that automation could pose to the company’s business, TaskUs had no shortage of other suitors for the massive growth equity round, according to one person familiar with the company. Indeed, Goldman Sachs and Softbank were among the other bidders for a piece of TaskUs, the source said.

Currently, the company has over 11,000 employees (including 2,000 in the U.S.) and is looking to expand.

“We chose to partner with Blackstone because they have a track record of building category defining businesses. Our goal is to build TaskUs into the world’s number one provider of tech enabled business services.  This partnership will help us dramatically increase our investment in consulting, technology and innovation to support our customer’s efforts to streamline and refine their customer experience,” said Maddock in a statement.

The transaction is expected to close in the fourth quarter of 2018, subject to regulatory approvals and customary closing conditions.