Category: UNCATEGORIZED

Clark, the tutor management business-in-a-box service, has been acquired by the New York-based education startup Noodle for an undisclosed amount, TechCrunch has learned.

Founded by John Katzman, the serial entrepreneur behind education technology giants including The Princeton Review and 2U, Noodle offers education search services to help people apply to the right programs that meet their needs.

Megan O’Connor, the co-founder and chief executive of Clark, actually met Katzman two weeks after she launched the company, which is backed by investors including Lightspeed Venture Partners, Winklevoss Capital, Rethink Education, Flat World Partners and Human Ventures (where O’Connor worked as the chief growth officer).

It’s not a stretch to call Katzman the godfather of tutoring, and, from the beginning, the seasoned executive took an interest in what Clark was doing, according to O’Connor.

With the acquisition, Clark’s shareholders will receive an equity stake in Noodle and O’Connor and her co-founder, Sam Gimbel, will take roles within Noodle to build out a tutoring service within the company, O’Connor says.

Going forward, Gimbel and O’Connor will build up the tutoring component of Noodle’s business as a complement to the company’s higher education and elementary and secondary school divisions.

One of the core components of the new tutoring platform within Noodle will be a focus on the individualization and personalization of tutoring sessions, buoyed by a community of tutors who share information on the most effective teaching strategies for different kinds of students.

What the tutoring practice won’t do, O’Connor says, is teach to a standardized curriculum. “If we can give them the software of shared services, then they can be more hands-on with the student,” O’Connor says.

Lyft is facing another lawsuit pertaining to its handling of alleged sexual assaults at the hands of drivers on its platform. In a suit filed today in the San Franciso Superior Court, Alison Turkos accuses Lyft of eleven counts, including general negligence, vicarious liability for assault with a deadly weapon, sexual assault, and sexual battery, and breach of contract.

The lawsuit describes how the plaintiff’s Lyft driver allegedly kidnapped her at gunpoint and took her across state lines, where the driver and other men took turns raping her, the lawsuits states.

“Alison remembers the men cheering and high fiving each other as they continued to rape her,” the lawsuit alleges. “Their attack was so brutal that the next day Alison experienced severe vaginal pain and bleeding. Her body was so exhausted from the attack and resulting trauma that Alison could not even leave her bed or raise her arms.”

When the plaintiff reported it to Lyft, the lawsuit alleges Lyft simply apologized for “inconvenience” and gave her a partial refund for the ride. The plaintiff says she reported the crime to the police, who performed a rape kit that found evidence of semen from at least two men on the clothing she wore that night.

The New York Police Department then transferred the case to the FBI, according to the lawsuit. The lawsuit states the FBI is now investigating the incident as a human trafficking case. However, Lyft “has been wholly uncooperative” throughout the NYPD and FBI’s investigation, the lawsuit alleges.

The lawsuit seeks special damages, including economic restitution to cover past and future hospital expenses, as well as expenses relating to her profession and loss of earning capacity.

“By failing to take reasonable steps to confront the problem of multiple rapes and sexual assaults of LYFT passengers by LYFT drivers, LYFT has acted in conscious disregard of the safety of its passengers,” the lawsuit alleges.

This suit comes just weeks after fourteen women filed suit against Lyft alleging the company has not addressed complaints pertaining to sexual assault. Both suits recommended Lyft adopt new policies, such as the addition of a surveillance camera to the app that can record audio and video of all rides.

Meanwhile, Lyft recently announced new safety features, including trip check-ins if a ride seems to be taking longer than it should and in-app 911 calling.

“We’re committed to playing a significant role in connecting our communities with transportation, and we understand the responsibilities that come along with that,” Lyft co-founder and President John Zimmer wrote in a blog post. “We’ve known since the beginning that as part of our mission, we must heavily invest in safety. We continue to welcome accountability and partnership to best protect our rider and driver community.”

It’s no coincidence that Lyft announced these safety features in light of the lawsuit on behalf of those fourteen women. The company had previously taken some steps to address safety, but at a much slower pace than competitor Uber, which has also faced a number of sexual assault and abuse lawsuits. Between 2014-2018, CNN found 103 Uber drivers who had been accused of sexual assault or abuse of passengers.

Over the years, both companies have taken steps to ramp up their respective safety procedures. In April, Uber launched a campus safety initiative while Lyft implemented continuous background checks and enhanced its identity verification process for drivers. Uber, however, implemented continuous background checks about a full year before Lyft, and added an in-app 911 calling feature more than a year before Lyft.

“We don’t take lightly any instances where someone’s safety is compromised, especially in the rideshare industry, including the allegations of assault in the news last week,” Zimmer said earlier this month in that same blog post. “The reality is that certain populations carry a disproportionate burden simply trying to get to work or back home after a night out — in the U.S., one in six women will face some form of sexual violence in their lives. The onus is on all of us to learn from any incident, whether it occurs on our platform or not, and then work to help prevent them.”

TechCrunch is awaiting comment from Lyft regarding this lawsuit. We’ll update this story if we hear back.

GoDaddy’s website-building product GoCentral is getting an upgrade today — and along with new features, there’s a new name: Websites+Marketing.

As you can probably guess, Websites+Marketing isn’t just a website builder. After all, as Senior Director of Product Management Heidi Gibson put it, a small business website is now part of a “a whole ecosystem that comprises your online presence.”

These are issues that Gibson said she’s experienced directly, as the chef/owner/”Commander in Cheese” at The American Grilled Cheese Kitchen in San Francisco.

“Our typical customer, our target customer is not just a small business — they’ve got one to five employees … they don’t know what they’re supposed to do, they don’t know what’s effective,” she added. Complicating matters is the fact that “where you need to be will not be the same answer for every kind of business.”

So GoDaddy Websites+Marketing — which Gibson described as “an evolution of GoCentral” — includes tools to manage email marketing and search engine optimization, and it syncs up with Facebook, Yelp, Instagram and Google My Business, so that it’s easy to read the latest reviews and comments, respond and post other updates directly from your Websites+Marketing dashboard.

It also includes a new feature called GoDaddy Insight, which relies on anonymized data — aggregated from all the businesses using GoDaddy and GoCentral — to provide entrepreneurs with a score on how their online presence and marketing compares to similar businesses, as well as an action plan recommending the next steps for improvement.

The website-builder looks pretty slick, too. Gibson acknowledged that some of the features will look pretty similar to anyone who’s used a competing product, but she said even here, GoDaddy has taken steps to make things easier.

For example, the Site Makeover feature allows businesses to get a quick view of how their content might look laid out on each of the 20-plus website templates, rather than making them click through each one. And thanks to GoDaddy’s recent acquisition of Sellbrite, businesses also can manage their product listings across online marketplaces like Amazon, Walmart and eBay.

GoDaddy Websites+Marketing is available in four pricing tiers, ranging from $10 to $25 per month.

A real check to Facebook CEO Mark Zuckerberg’s control is finally coming in the form of a 11- to 40-member Oversight Board that will review appeals to its policy decisions like content takedowns and make recommendations for changes. Today Facebook released the charter establishing the theoretically independent Oversight Board, with Zuckerberg explaining that when it takes a stance, “The board’s decision will be binding, even if I or anyone at Facebook disagrees with it.”

Slated to be staffed up with members this year who will be paid by a Facebook established trust (the biggest update to its January draft charter), the Oversight Board will begin judging cases in the first half of 2020. Given Zuckerberg’s overwhelming voting control of the company, and the fact that its board of directors contains many loyalists like COO Sheryl Sandberg and investor Peter Thiel who he’s made very rich, the Oversight Board could ensure the CEO doesn’t always have the final say in how Facebook works.

But in some ways, the committee could serve to shield Zuckerberg and Facebook from scrutiny and regulation much to their advantage. The Oversight Board could remove total culpability for policy blunders around censorship or political bias from Facebook’s executives. It also might serve as a talking point towards the FTC and other regulators investigating it for potential anti-trust violations and other malpractice, as the company could claim the Oversight Board means it’s not completely free to pursue profit over what’s fair for society.

Finally, there remain serious concerns about how the Oversight Board is selected and the wiggle room the charter provides Facebook. Most glaringly, Facebook itself will choose the initial members and then work with them to select the rest of the board, and thereby could avoid adding overly incendiary figures. And it maintains that “Facebook will support the board to the extent that requests are technically and operationally feasible and consistent with a reasonable allocation of Facebook’s resources”, giving it the right to decide if it should apply the precedent of Oversight Board verdicts to similar cases or broadly implement its policy guidance.

How The Oversight Board Works

When a user disagrees with how Facebook enforces its policies, and with the result of an appeal to Facebook’s internal moderation team, they can request an appeal to the oversight board. Examples of potential cases include someone disagreeing with Facebook’s refusal to deem a piece of content as unacceptable hate speech or bullying, its choice to designate a Page as promoting terrorism and remove it, or the company’s decision to leave problematic content such as nudity up because its newsworthy. Facebook can also directly ask the Oversight Board to review policy decisions or specific cases, especially urgent ones with real-world consequences.

After Zuckerberg initially laid out a blueprint for the Oversight Board a year ago, Facebook assigned a 100-person team to build out the plan for the board. It held 6 workshops and 22 round-tables plus case review simulations with 650 people from 88 countries.

The board will include a minimum of 11 members but Facebook is aiming for 40. They’ll serve three year terms and a maximum of three terms each as a part-time job, with appointments staggered so there isn’t a full change-over at any time. Facebook is looking for members with a broad range of knowledge, competencies, and expertise who lack conflicts of interest. They’re meant to be “experienced at deliberating thoughtfully and collegially”, “skilled at making and explaining decisions based on a set of policies”, “well-versed on matters relating to digital content and governance”, and “independent and impartial”. 

Facebook will appoint a set of trustees that will work with it to select initial co-chairs for the board, who will then assist with sourcing, vetting, interviewing, and orienting new members. The goal is “broad diversity of geographic, gender, political, social and religious representation”. The trust, funded by Facebook with an as yet undecided amount of capital, will set members’ compensation rate in the near future and oversee term renewals.

Inevitable Calls Of Biased Board Members

My biggest worry here is how Facebook will handle the fact that it’s trying to represent an extraordinarily vast set of global policy perspectives…broader than any one country’s laws. What’s taboo or even illegal in one nation may be common or lauded in another. Facebook may see endless challenges from different segments of the public regarding the previous public statements by board members.

What Facebook’s own staff in California might see as an uncontroversial view point could trigger calls for removal from the board elsewhere. We’ve seen how common “cancelled” culture has become when the public digs up problematic content from celebrities or politicians, and that’s just based on what flies in the United States.

For example, Republican Senators just bullied Facebook into removing a fact-check that found the statement “abortion is never medically necessary” to be false, allowing that view point to spread uninhibited on the social network. I personally wouldn’t want someone with that view point on the Oversight Board, but others might feel the opposite. And what happens when politicians start demanding more conservative representation on the Oversight Board the same way they’ve badgered Facebook for supposedly censoring them despite evidence to the contrary?

What Cases Get Reviewed

The board will choose which cases to review based on their significance and difficulty. They’re looking for issues that are severe, large-scale and important for public discourse, while raising difficult questions about Facebook’s policy or enforcement that is disputed, uncertain, or represents tension or trade-offs between Facebook’s recently codified values of authenticity, safety, privacy, and dignity. The board will then create a sub-panel of five members to review a specific case.

The board will be able to question that request that Facebook provide information necessary to rule on the case with a mind to not violating user privacy. They’ll interpret Facebook’s Community Standards and policies and then decide whether Facebook should remove or restore a piece of content and whether it should change how that content was designated. Verdicts are meant to have consensus but will be approved by majority when necessary.

How Decisions Get Made

Once a panel makes a draft decision, it’s circulated to the full board who can recommend a new panel review it if a majority take issue with the verdict. Once they’ve gone through a privacy review to protect the identities of those involved with the case, the decisions will be made public within two weeks and affected users will be notifed. Those decisions will be archived in a database, and are meant to act as precedent for future decisions. The idea is that the decisions of the board will be binding and implemented by Facebook as long as they don’t require it to violate the law.

But Will Facebook Really Implement Them?

The biggest concern with the charter is that it still provides Facebook some leeway about how to implement the board’s decisions. Critically, it only has to apply the decision to the specific case reviewed, and it’s at the company’s discretion to turn that into blanket policy.

“In instances where Facebook identifies that identical content with parallel context — which the board has already decided upon — remains on Facebook, it will take action by analyzing whether it is technically and operationally feasible to apply the board’s decision to that content as well. When a decision includes policy guidance or a policy advisory opinion, Facebook will take further action by analyzing the operational procedures required to implement the guidance, considering it in the formal policy development process of Facebook . . . Facebook will support the board to the extent that requests are technically and operationally feasible and consistent with a reasonable allocation of Facebook’s resources.“

Because of these sections I’ve bolded, Facebook has the ability to decide it would be operationally infeasible to do what the board decided in every situation, merely take the guidance into account for future policy-making, and choose whether implementation is a reasonable allocation of capital and staff. This provides a sizable grey area.

If Facebook chooses that the board’s decision could materially reduce sharing even if it protected users, it might considering that operationally infeasible. If it would cost too much to moderate content in the way the board recommends, it could deem that unreasonable resource allocation. And if the policy guidance doesn’t mesh with its other ojbectives, it only has to “consider” the board’s wishes.

This section is where advocates and critics should focus. These exemptions to implementation need to be made less vague if the structure is truly going to hold Facebook accountable. If Facebook just declines to broadly change its policy to fit the board’s recommendation, all the board can do is make binding decisions on specific cases.

Facebook director of governance Brent Harris explained on a call with reporters that “If the board doesn’t feel like we’ve handled it right, they’ll keep taking cases and overturn us.” But again the board’s power is focused on a case-by-case basis. Facebook still controls the wide-reaching changes to policy.

If you want to learn more about solutions to Facebook’s concentration of power, check out my talk at SXSW 2020 with Facebook’s co-founder Chris Hughes who has called for the company to be broken up.

Amazon-owned Twitch has made a small but strategic acquisition designed to improve its search capabilities and better direct viewers to exactly the right content. The company is acquiring IGDB, the Internet Games Database (no relation to Amazon’s IMDb), a website dedicated to combining all the relevant information about games into a comprehensive resource for gamers everywhere. As a result of the acquisition, IGDB’s database will now feed into Twitch’s search and discovery feature set. However, the IGDB website itself will not be shut down.

Founded in 2015 by Christian Frithiof and a small team based in Gothenburg, Sweden, IGDB sources its gaming content both through community contributions and automation.

The site includes useful information for every game, like the genre, platforms supported, description, member and critic ratings and reviews, storyline, game modes, publisher, release dates, characters, and more. You could also find less common details like how long it would take to play the game in question, or the player perspectives the game offered, among other things.

And similar to IMDb’s mission of organizing everything associated with the entertainment industry, IGDB allowed voice talent to claim their profile on its site, in addition to listing the full credits associated with a given title.

To generate revenue, IGDB provided a developer API that’s been free to use for smaller shops or $99 per month for up to 50K requests. Interested partners, (e.g. ASUS), could reach out to request special pricing. To date, IGDB was working with several thousand API users, we understand. 

Twitch confirmed the acquisition to TechCrunch in a statement.

“Millions of people come to Twitch every day to find and connect with their favorite streamers and communities, and we want to make it easier for people to find what they’re looking for,” a spokesperson said. “IGDB has developed a comprehensive gaming database, and we’re excited to bring them on to help us more quickly improve and scale search and discovery on Twitch.”

Deal terms were not disclosed, but it was likely a small deal, from a financial standpoint. IGDB is only a 10-person team and had raised just $1.5 million to date, according to data from Crunchbase.

From a strategic standpoint, however, the acquisition is much more impactful.

Twitch CEO Emmett Shear has spoken publicly about the issues surrounding Twitch’s search functionality and how it needs to improve on that front.

“We want every place on Twitch to help you get discovered. Today, nearly one in three people who come to Twitch use Search to find what they’re looking for. Now, I’ll be the first to admit that our search function hasn’t always been the best experience,” Shear had said earlier this year, speaking at TwitchCon Berlin.”One wrong letter and your search results may come back empty, or direct you to a very different streamer than the one you were looking for. So we’re going to fix search so it actually works,” he promised.

In recent weeks, there were hints that something was going on at IGDB.

In a blog post dated August 19, 2019, IGDB announced it was starting a “large scale migration of our backend, database, and hosting” and said that the service was “about to undergo some changes, some temporary and others more permanent.” As a part of its changes, it shut off the ability for users to sign-up or update their profiles, and it shut down its pulse news, feed, and recommendations features.

Now a part of Twitch, IGDB will merge its free and premium APIs into one free tier, will clean up other features, and migrate infrastructure. Its IGDB website will continue to remain online.

“Our mission has always been to build the most comprehensive gaming database in the world. Such a monumental undertaking can be quite challenging when you are a small startup team,” reads an IGDB blog post. “By joining Twitch, we will be able to tap into their experience, resources, and skills, which will enable us to accelerate our progress and deliver the version of IGDB we all always dreamed about. Not only that, our companies share the same culture, core values, and passion for gaming– making this the perfect fit,” the post said.

It was common industry knowledge Twitch previously used competing data provider Giantbomb. As is often the case, the company may have been in discussions with IGDB about making switch which led to the acquisition. (The company declined to say how it can about.) What had made IGDB different from other API providers, like Mobygames, is that it allowed its API to be used commercially, including by competing projects, and it allowed caching and storing data on local databases.

The entire 10-person team from IGDB will remain based in Sweden, but will report into Twitch through its Viewer Experience organization.

 

 

 

Colorcon, a 58-year-old company that develops, supplies, and supports specialty products for the pharmaceutical industry — think food colorants, nutritional coatings, the film on time-release medications — is getting into the business of funding startups.

It isn’t bringing aboard any traditional venture investors or taping off a corner of its Harleysville, Pa., offices so a team of staffers can meet with startups. Colorcon knows what it doesn’t know, suggests its CEO, Martti Hedman. “We’re a team of 750 people, so we didn’t want to invest in our own VC team. We’re sort of too small for that, too inexperienced.”

Instead, Colorcon, which wants to plug $50 million into startups, has elected to outsource its venture operations to a low-flying but growing outfit in San Francisco called Touchdown Ventures that helps manage the corporate venture activities of a dozen companies already, including the multinational food company Kellogg’s, the media giant 21st Century Fox, and the adhesive manufacturing company Avery Dennison.

The idea is for Touchdown, which has 30 employees, to use its relationships with these other companies and with VCs — along with its own outbound efforts and sector analyses — to  bring to Colorcon deals it might want to fund. After that, Hedman, Colorcon CFO Dave Graeber, and Colorcon’s head of corporate development, Pankaj Rege, will decide which startups merit checks.

The initial amount it intends to commit: between $1.5 million and $3 million per startup, says Touchdown Ventures CEO David Horowitz, who will be heading up the effort for Colorcon.

It’s worth noting that unlike many companies that work with pharmaceutical giants, Colorcon isn’t looking to fund startups that are developing active ingredients or molecules. Instead, the fund will target investments across the manufacturing, supply chain, and delivery of pharmaceutical products and services.

As you might imagine, it’s also looking to invest in startups where it can add value, be it subject matter expertise or introductions to the many companies and labs with which it works around the world. Indeed, more than 70 percent of Colorcon’s sales comes outside of North America, including in China, India, South America, and, to a small but growing extent, Europe.

Is it a match made in heaven? We’ll see. But it’s certainly interesting to see companies like Colorcon with their industry expertise hitching their wagons to platforms like Touchdown, which has institutional know-how about VC.

Says Horowitz, “We have other heath care relationships. We know how to drum up deal flow. We have relationships with thousands of VCs and we speak at conferences and we do things that are specific to this industry.” That work now produces 5,000 “opportunities” per year, in terms of startup pitches, Horowitz adds.

If Colorcon is lucky, some of those startups — whether Colorcon acquires them or adopts their technology or merely learns from them —  will help keep the company in business for another 50 years.

French president Emmanuel Macron announced in a speech ahead of France Digitale Day that the French government has convinced institutional investors to invest more heavily in late-stage VC funds and asset managers in one way or another. Institutional investors have committed to investing $5.5 billion (€5 billion).

“We’ll have €2 billion that will go in so-called late-stage funds and €3 billion for funds managed by asset managers specialized in tech,” Macron said.

In addition to that financial pledge, the French government wants to break down any hurdle that prevents French startups from raising $100 million+ funding round in France, becoming a unicorn and eventually going public.

A couple of years ago, Macron gave a speech at Viva Technology in Paris. It was the first time he addressed the startup community after his election. At the time, I wrote: “Macron wanted to send a message to the startup community — he still cares about technology very much, thank you for asking.”

Since then, the French tech ecosystem has thrived, but without any radical policy change to shake things up. But today marks a departure as it’s all about startups, startups and startups.

It’s clear that Macron believes that startups represent a huge opportunity when it comes to job creation, competitiveness and reshaping the economic landscape in France. In other words, according to him, if you help startups thrive, it’s going to trickle down all the way and have positive impacts on your neighbor who has never used a computer in her life.

Some will applaud such a move, others will say that it divides society.

“When I talk about startup funding, I talk about the ability to help those startups succeed,” Macron said. “I’m talking about the jobs of tomorrow. And I’m saying that for many French citizens who think that those are only financial numbers.”

Financing hypergrowth

So here’s Macron’s plan. First, French VC funds have been good when it comes to funding startups at the seed, Series A and sometimes Series B level. But many startups then look for international investors for late-stage rounds. For instance, just last week, Akeneo raised $46 million in a round led by Summit Partners, a Boston-based VC firm.

“Numbers show that we’re getting there, and I want to start from there,” Macron said. “The goal when it comes to technology is that we should be one of the countries that matter. Fundraising from French startups keep setting new records — we had $3.1 billion in fundraising in 2017, $4 billion in 2018 and $5.5 billion in 2019 probably.”

Following a report from Philippe Tibi, the French government has been working on a way to foster late-stage funds and investments in public tech companies in France. “We managed to rally big insurance companies, asset managers and long-term public investment funds,” a source close to Macron told me.

Private companies, such as Axa, Generali and Allianz, as well as public investors, such as EDF, Caisse des Dépôts, the pension reserve fund, are all going to invest in late-stage VC. Overall, two-thirds of them are private companies, one-third of them are public institutions, according to the source.

They’ll have three ways to invest and take part in the initiative:

• If they have their own VC fund, they can create a new late-stage fund.
• If they are limited partners in various VC funds, they can invest in late-stage funds managed by third-party teams.
• If they don’t know anything about venture capital, they can invest in a special fund of funds managed by Bpifrance. Bpifrance will then select various late-stage funds and invest that money in those funds.

Eventually, the French government hopes that there will be at least 10 French VC firms with a late-stage fund above €1 billion. By pushing them to redirect some of their investments in VC, the French government thinks that they’ll invest more regularly in venture capital in the future.

When it comes to going public, the French government wants to make European stock exchanges more attractive. They're hoping the new influx of late-stage cash will convince banks and other financial institutions that manage huge positions in tech companies to create local teams in Paris.

Attracting foreign VCs too

French startups still want to become global players and the French government is well aware of that. And foreign VCs shouldn’t be at odds with French VC firms.

That’s why the French government also invited around 40 partners of venture capital firms and limited partners for a couple of days in Paris this week. They’ll meet key people in the ecosystem as well as promising startups.

I covered the first edition of this tour last year. The message was clear: Foreign VC firms should think about investing in French startups. Some are already doing it while others never thought about it. And the thing is nobody wants to be the first one to invest in something new, but nobody wants to be the last one, either.

This year, the French government is inviting a new batch of foreign investors from Khosla Ventures, Accel, Andreessen Horowitz, etc. There are more Asian investors in the mix this time round.

But Macron said that France should control its own destiny when it comes to startup funding. “When I talk about sovereignty, I deeply believe in that concept. It’s a politically-charged word, but I think it’s at the heart of your approach. I believe in technological and economical sovereignty,” Macron said.

Transforming La French Tech

The French Tech Mission, also known as La French Tech, is a government-backed initiative that promotes French startups around the world and provides a few services to help startups.

And the government is going to overhaul the French Tech Mission drastically. This is as significant as the late-stage funding news. In addition to the small core team, every French ministry and administration will have a French Tech correspondent — Urssaf, INPI, AFNOR, Banque de France, customs, etc. Eventually, there will be 150 people spread out across the entire government working in some way or another for French startups.

“We’re not alone, we get to coordinate with everyone,” French Tech Mission director Kat Borlongan told me. “The overarching announcement is that France is going all in.”

La French Tech is going to become a one-stop shop for tech startups to overcome any administrative hurdle. La French Tech is going to pick 40 (and later 120) top-performing startups and give them the label Next40 and French Tech 120 — a play on words with the CAC40 and SBF 120 stock indexes. Those companies will automatically be able to access this fast-track administrative system — every startup will get a representative for their particular needs. This special treatment proves that startups have become a center piece of France’s economic policies.

“The coolest thing is that they can ask us for anything: ‘I’m about to do bizdev in China’, ‘I’m launching a rocket and I need to test it on a space facility’ or ‘I’m hiring 50 people and I need them and all their families here’,” Borlongan told me.

All companies that are unicorns or have raised more than €100 million are automatically in the Next40. Then, the government is looking at growth rate and annual turnover to find the most promising 40 and 120 startups.

“I’ll leave you with a goal: there should be 25 [French] unicorns by 2025,” Macron said at the end of his speech.

The Daily Crunch is TechCrunch’s roundup of our biggest and most important stories. If you’d like to get this delivered to your inbox every day at around 9am Pacific, you can subscribe here.

1. Review: The iPhone 11 Pro and iPhone 11 do Disneyland after dark

Matthew Panzarino continues his tradition of testing out the latest iPhones at Disneyland. This time, he was particularly interested in how well the iPhone 11’s Night Mode works. His verdict: It compares extremely well to other low-light cameras, with exposure and color rendition that’s best-in-class.

But if you’re planning to upgrade, should you get the Pro, or the regular ol’ iPhone 11? Apparently the Pro is really there to address edge cases — the best video and photo options, a better dark mode experience, a brighter screen.

2. Under pressure, The We Company now only says it expects to go public ‘by the end of the year’

A new note from WeWork’s parent company all but confirms that it is indeed delaying its IPO roadshow, which had been expected to commence this week.

3. Amazon launches Amazon Music HD with lossless audio streaming

Amazon has a new, high-quality streaming tier of its music service called Amazon Music HD. It’s priced at $12.99 per month for Prime members, and you can add it to your existing Amazon Music subscription for an additional $5 each month.

4. Will Smith and Ang Lee are coming to Disrupt SF

They’ll be joining us to discuss their upcoming film “Gemini Man,” which features “jaw-dropping effects” from Weta Digital. The effects allow Smith to play both an assassin named Henry Brogan and a younger clone who’s been sent to kill his older counterpart.

5. Computer scientist Richard Stallman, who defended Jeffrey Epstein, resigns from MIT CSAIL and the Free Software Foundation

Stallman said he has resigned from his position as a visiting scientist at MIT’s Computer Science and Artificial Intelligence Lab after describing a victim of sex trafficker Jeffrey Epstein as “entirely willing” in emails sent to a department list.

6. I hope Apple Arcade makes room for weird, cool shit

Apple Arcade seems purpose-built to make room in the market for beautiful, sad, weird, moving, slow, clever and heartfelt.

7. What startup CSOs can learn from three enterprise security experts

How do you keep your startup secure? That’s one of the big questions we explored at TC Sessions: Enterprise earlier this month — and if you weren’t there, we’ve got a write-up of the main takeaways. (Extra Crunch membership required.)

The latest research from OpenAI put its machine learning agents in a simple game of hide-and-seek, where they pursued an arms race of ingenuity, using objects in unexpected ways to achieve their goal of seeing or being seen. This type of self-taught AI could prove useful in the real world as well.

The study intended to, and successfully did look into the possibility of machine learning agents learning sophisticated, real-world-relevant techniques without any interference of suggestions from the researchers.

Tasks like identifying objects in photos or inventing plausible human faces are difficult and useful, but they don’t really reflect actions one might take in a real world. They’re highly intellectual, you might say, and as a consequence can be brought to a high level of effectiveness without ever leaving the computer.

Whereas attempting to train an AI to use a robotic arm to grip a cup and put it in a saucer is far more difficult than one might imagine (and has only been accomplished under very specific circumstances); the complexity of the real, physical world make purely intellectual, computer-bound learning of the tasks pretty much impossible.

At the same time, there are in-between tasks that do not necessarily reflect the real world completely, but still can be relevant to it. A simple one might be how to change a robot’s facing when presented with multiple relevant objects or people. You don’t need a thousand physical trials to know it should rotate itself or the camera so it can see both, or switch between them, or whatever.

OpenAI’s hide-and-seek challenge to its baby ML agents was along these lines: A game environment with simple rules (called Polyworld) that nevertheless uses real-world-adjacent physics and inputs. If the AIs can teach themselves to navigate this simplified reality, perhaps they can transfer those skills, with some modification, to full-blown reality.

Such is the thinking behind the experiment, anyway, but it’s entertaining enough on its own. The game pits two teams against one another in a small 3D arena populated with a few randomly generated walls and objects. Several agents are spawned in it and the “hiders” are given a few seconds to familiarize themselves with the environment and hide. They can also interact with the objects, pushing them around and locking them in place. Once the “seeker” looks up they have a certain amount of time to spot the hiders.

All the machine learning program was informed of were the basic senses — each agent has a view of the world and a sort of “lidar” to tell them the positions of nearby objects — and the knowledge that they can move objects around. But beyond that they were simply given a point when they succeeded at their job, either seeking or hiding — that’s their motivation.

From these basic beginnings came wonderfully interesting results. At first the teams essentially moved randomly. But over millions of games the hiders found that by moving in a certain way — “crudely” running away — they could increase their points more often. And thus the arms race began.

Since the hiders were running away, the seekers began to develop more structured pursuit behaviors. But then came the tools.

Hiders found that they could push and pull objects on the field, and began constructing forts out of them.

The seekers responded by using ramps found on the field to ascend over the objects and crash the forts.

Hiders began looking for ramps while the seekers were still counting and either stealing them or locking them in place (the lock can’t be undone by an agent of the opposite team). Interestingly this involved a certain amount of division of labor, because there often wasn’t enough time for both hiders to go out in search of ramps; one had to prepare the fort while the other went on ramp patrol.

The OpenAI team thought maybe that was the end of it, but the seekers learned to exploit an unintended feature of the game world. They discovered that they could “push” a box while riding on top of it, meaning they could put it near a ramp, climb onto it, then “surf” it all over the field and find those naughty hiders.

Of course the hiders responded by locking every object they weren’t using to construct their fort — and that seems to be the end of the line as far as strategy in this game.

So what’s the point? As the authors of the paper explain, this is kind of the way we came bout.

The vast amount of complexity and diversity on Earth evolved due to co-evolution and competition between organisms, directed by natural selection. When a new successful strategy or mutation emerges, it changes the implicit task distribution neighboring agents need to solve and creates a new pressure for adaptation. These evolutionary arms races create implicit autocurricula whereby competing agents continually create new tasks for each other.

Inducing autocurricula in physically grounded and open-ended environments could eventually enable agents to acquire an unbounded number of human-relevant skills.

In other words, having AI models compete in an unsupervised manner may be a far better way to develop useful and robust skills than letting them toddle around on their own, racking up an abstract number like percentage of environment explored or the like.

Increasingly it is difficult or even impossible for humans to direct every aspect of an AI’s abilities by parameterizing it and controlling the interactions it has with the environment. For complex tasks like a robot navigating a crowded environment, there are so many factors that having humans design behaviors may never produce the kind of sophistication that’s necessary for these agents to take their place in everyday life.

But they can teach each other, as we’ve seen here and in GANs, where a pair of dueling AIs work to defeat the other in the creation or detection of realistic media. The OpenAI researchers posit that “multi-agent autocurricula,” or self-teaching agents, are the way forward in many circumstances where other methods are too slow or structured. They conclude:

“These results inspire confidence that in a more open-ended and diverse environment, multi-agent dynamics could lead to extremely complex and human-relevant behavior.”

Some parts of the research have been released as open source. You can read the full paper describing the experiment here.