Elon Musk says SpaceX to double launch pad usage for Starship tests, Super Heavy flights coming in a ‘few months’

SpaceX is set to significantly ramp up its Starship development program in the new year, in more ways than one. SpaceX CEO and founder Elon Musk noted on Twitter on Thursday that the company will seek to make use of both of its two launch pads at its development facility in Boca Chica, Texas with prototype rockets set up on each, and that it will begin flight testing its Super Heavy booster (starting with low-altitude “hops”) as quickly as “a few months” from now.

Recently, SpaceX set up its SN9 prototype of Starship (the ninth in the current series) at Pad B at its Texas testing facility, which is on the Gulf of Mexico. SN9 will be next to undergo active testing, after SpaceX successfully flew its predecessor SN8 to an altitude of around 40,000 feet, and then executed a crucial belly flop maneuver that will be used to help control the powered landing of the production version. SN8 was destroyed when it touched down harder than expected, but SpaceX still achieved all its testing goals with the flight — and more.

SN9 will now undergo ground tests before hopefully doing its own flight test later on. That’ll provide the team with even more valuable data to carry on to further tests — with the ultimate goal of eventually achieving orbit with a Starship prototype vehicle. Musk’s tweet that two prototypes will be stood up next to each other on Pad A and Pad B at the Boca Chica site could indicate the pace of these test flights might speed up, to match the fast clip at which SpaceX is constructing new rocket iterations.

Meanwhile, news that Super Heavy could be undergoing testing soon is also reason to get excited about 2021 for SpaceX and Starship. Super Heavy is the booster that SpaceX will eventually use to fly Starship for orbital launches, and to eventually help propel it to deep space — for destinations including Mars. Super Heavy will be around 240 feet tall, and will include 28 Raptor engines to provide it with the lift capacity needed to break Earth’s gravity when it’s stacked with a Starship loaded down with cargo.


Source: Tech Crunch

Use Git data to optimize your developers’ annual reviews

The end of the year is looming and with it one of your most important tasks as a manager. Summarizing the performance of 10, 20 or 50 developers over the past 12 months, offering personalized advice and having the facts to back it up — is no small task.

We believe that the only unbiased, accurate and insightful way to understand how your developers are working, progressing and — last but definitely not least — how they’re feeling, is with data. Data can provide more objective insights into employee activity than could ever be gathered by a human.

It’s still very hard for many managers to fully understand that all employees work at different paces and levels.

Consider this: Over two-thirds of employees say they would put more effort into their work if they felt more appreciated, and 90% want a manager who’s fair to all employees.

Let’s be honest. It’s hard to judge all of your employees fairly if you’re (1) unable to work physically side-by-side with them, meaning you’ll inevitably have more contact with the some over others (e.g., those you’re more friendly with); and (2) you’re relying on manual trackers to keep on top of everyone’s work, which can get lost and take a lot of effort to process and analyze; (3) you expect engineers to self-report their progress, which is far from objective.

It’s also unlikely, especially with the quieter ones, that on top of all that you’ll have identified areas for them to expand their talents by upskilling or reskilling. But it’s that kind of personal attention that will make employees feel appreciated and able to progress professionally with you. Absent that, they’re likely to take the next best job opportunity that shows up.

So here’s a run down of why you need data to set up a fair annual review process; if not this year, then you can kick-start it for 2021.

1. Use data to set next year’s goals

The best way to track your developers’ progress automatically is by using Git Analytics tools, which track the performance of individuals by aggregating historical Git data and then feeding that information back to managers in minute detail.

This data will clearly show you if one of your engineers is over capacity or underworked and the types of projects they excel in. If you’re assessing an engineering manager and the team members they’re responsible for have been taking longer to push their code to the shared repository, causing a backlog of tasks, it may mean that they’re not delegating tasks properly. An appropriate goal here would be to track and divide their team’s responsibilities more efficiently, which can be tracked using the same metrics, or cross-training members of other teams to assist with their tasks.

Another example is that of an engineer who is dipping their toe into multiple projects. Indicators of where they’ve performed best include churn (we’ll get to that later), coworkers repeatedly asking that same employee to assist them in new tasks and of course positive feedback for senior staff, which can easily be integrated into Git analytics tools. These are clear signs that next year, your engineer could be maximizing their talents in these alternative areas, and you could diversify their tasks accordingly.

Once you know what targets to set, you can use analytics tools to create automatic targets for each engineer. That means that after you’ve set it up, it will be updated regularly on the engineer’s progress using indicators directly from the code repository. It won’t need time-consuming input from either you or your employee, allowing you both to focus on more important tasks. As a manager you’ll receive full reports once the deadline of the task is reached and get notified whenever metrics start dropping or the goal has been met.

This is important — you’ll be able to keep on top of those goals yourself, without having to delegate that responsibility or depend on self-reporting by the engineer. It will keep employee monitoring honest and transparent.

2. Three Git metrics can help you understand true performance quality

The easiest way for managers to “conclude” how an engineer has performed is by looking at superficial output: the number of completed pull requests submitted per week, the number of commits per day, etc. Especially for nontechnical managers, this is a grave but common error. When something is done, it doesn’t mean it’s been done well or that it is even productive or usable.

Instead, look at these data points to determine the actual quality of your engineer’s work:

  1. Churn is your number-one red flag, telling you how many times someone has modified their code in the first 21 days after it has been checked in. The more churn, the less of an engineer’s code is actually productive, with good longevity. Churn is a natural and healthy part of the software development process, but we’ve identified that any churn level above the normal 15%-30% indicates that an engineer is struggling with assignments.


    Source: Tech Crunch

US seed-stage investing flourished during pandemic

As the United States entered its first wave of COVID-19 lockdowns, there were wide expectations in startup land that a reckoning had arrived. But the expected comeuppance of high-burn, high-growth startups fueled by cheap capital provided by venture capitalists raising ever-larger funds, failed to arrive.

Instead, the very opposite came to pass.

Layoffs happened swiftly and aggressively during the early months of the pandemic era. But by the middle of Q2, venture activity had warmed and third quarter dealmaking felt swift and competitive, with some investors describing it as the hottest summer in recent years.

Venture capital as an asset class has survived the pandemic’s stress test.

But somewhat lost amongst the splashy megarounds and high-interest IPOs that can dominate the news cycle were seed-stage startups. The raw little companies that represent the grist that will shape itself into the next set of giants.

TechCrunch explored what happened in seed investing to uncover what was missed amidst the storm and fury of late-stage startup activity. According to a TechCrunch analysis of PitchBook data and a survey of venture capitalists, a few trends became clear.

First, the pattern of rising seed-check sizes seen in prior years continued despite the tumultuous business climate. Second, more expensive and larger seed deals were not only caused by excessive capital present in the private markets. Instead, COVID-19 shook up which startups were considered attractive by private investors. And the changeup did not necessarily raise their number.

Let’s dig into the data and see what it can teach us about this wild year. Then we’ll hear from Eniac VenturesNihal Mehta, Freestyle’s Jenny Lefcourt, Pear VC’s Mar Hershenson and Contrary Capital’s Eric Tarczynski about what they saw in 2020 while writing a chunk of the checks that our data encompasses.

The American seed market in 2020

If you didn’t think much about seed in 2020, you’re not alone. Late, huge rounds consumed most of the media’s oxygen, leaving smaller startups to compete for scraps of attention. There was so much late-stage activity — around 90 $100 million or larger rounds in Q3, for example — it was difficult for smaller investments to command attention.

But despite living in the background, the dollars invested into seed-stage startups in the United States had an up-and-down year that was fascinating:

Image Credits: PitchBook

Seed dollar volume fell as Q1 progressed, reaching a 2020 nadir in April, the start of Q2. But as May arrived, the pace at which investors put money into seed-stage startups accelerated, recovering to January levels — which is to say, pre-pandemic — by June. The COVID dip, for seed, then, was a short-term affair.


Source: Tech Crunch

National Grid sees machine learning as the brains behind the utility business of the future

If the portfolio of a corporate venture capital firm can be taken as a signal for the strategic priorities of their parent companies, then National Grid has high hopes for automation as the future of the utility industry.

The heavy emphasis on automation and machine learning from one of the nation’s largest privately held utilities with a customer base numbering around 20 million people is significant. And a sign of where the industry could be going.

Since its launch, National Grid’s venture firm, National Grid Partners, has invested in 16 startups that featured machine learning at the core of their pitch. Most recently, the company backed AI Dash, which uses machine learning algorithms to analyze satellite images and infer the encroachment of vegetation on National Grid power lines to avoid outages.

Another recent investment, Aperio, uses data from sensors monitoring critical infrastructure to predict loss of data quality from degradation or cyberattacks.

Indeed, of the $175 million in investments the firm has made, roughly $135 million has been committed to companies leveraging machine learning for their services.

“AI will be critical for the energy industry to achieve aggressive decarbonization and decentralization goals,” said Lisa Lambert, the chief technology and innovation officer at National Grid and the founder and president of National Grid Partners.

National Grid started the year off slowly because of the COVID-19 epidemic, but the pace of its investments picked up and the company is on track to hit its investment targets for the year, Lambert said.

Modernization is critical for an industry that still mostly runs on spreadsheets and collective knowledge that has locked in an aging employee base, with no contingency plans in the event of retirement, Lambert said. It’s that situation that’s compelling National Grid and other utilities to automate more of their business.

“Most companies in the utility sector are trying to automate now for efficiency reasons and cost reasons. Today, most companies have everything written down in manuals; as an industry, we basically still run our networks off spreadsheets, and the skills and experience of the people who run the networks. So we’ve got serious issues if those people retire. Automating [and] digitizing is top of mind for all the utilities we’ve talked to in the Next Grid Alliance.

To date, a lot of the automation work that’s been done has been around basic automation of business processes. But there are new capabilities on the horizon that will push the automation of different activities up the value chain, Lambert said.

“ ML is the next level — predictive maintenance of your assets, delivering for the customer. Uniphore, for example: you’re learning from every interaction you have with your customer, incorporating that into the algorithm, and the next time you meet a customer, you’re going to do better. So that’s the next generation,” Lambert said. “Once everything is digital, you’re learning from those engagements — whether engaging an asset or a human being.”

Lambert sees another source of demand for new machine learning tech in the need for utilities to rapidly decarbonize. The move away from fossil fuels will necessitate entirely new ways of operating and managing a power grid. One where humans are less likely to be in the loop.

“In the next five years, utilities have to get automation and analytics right if they’re going to have any chance at a net-zero world — you’re going to need to run those assets differently,” said Lambert. “Windmills and solar panels are not [part of] traditional distribution networks. A lot of traditional engineers probably don’t think about the need to innovate, because they’re building out the engineering technology that was relevant when assets were built decades ago — whereas all these renewable assets have been built in the era of OT/IT.”

 


Source: Tech Crunch

Five VCs discuss what surprised them the most in 2020

Hello and welcome back to Equity, TechCrunch’s venture capital-focused podcast (now on Twitter!), where we unpack the numbers behind the headlines.

Today is our holiday look-back at the year, bringing not only our own Danny and Natasha and Chris and Alex into the mix, but also five venture capitalists who we got to leave us their notes as well. The goal for this episode was to reflect on a year that no one could have ever predicted, but with a specific angle, as always, on venture capital and startups.

We asked about the biggest surprise, non-portfolio companies to watch, and trends they got wrong and right. There was also banter on Zoom investing (Alex came up with Zesting, but we’re taking suggestions if anyone comes up with a better moniker) and startup pricing.

Here’s who we asked to call into our super Fancy Equity Hotline:

Thanks to them all for participating, and of course you, our dear Equity listeners, for a blockbuster year for the podcast.

Equity drops every Monday at 7:00 a.m. PST and Thursday afternoon as fast as we can get it out, so subscribe to us on Apple PodcastsOvercastSpotify and all the casts.


Source: Tech Crunch

Honk introduces a real-time, ephemeral messaging app aimed at Gen Z

A new mobile app called Honk aims to make messaging with friends a more interactive, real-time experience. Instead of sending texts off into the void and hoping for a response, friends on Honk communicate via messages that are shown live as you type, with no saved chat history and no send button. The end result is a feeling of being more present in a conversation, as Honk will notify users the moment someone leaves a chat. And if you really want to get someone’s attention, you can send them a “Honk” — a hard-to-miss notification to join your chat.

If it’s even more urgent, you can even spam the Honk button by pressing it repeatedly. This sends notifications to a friend’s phone if they’re off the app, or a flood of colorful emoji if they have the app open.

After setting up an account by customizing your profile pic, selecting a username and adding friends, you can then tap on a friend’s name in your list to send them a message.

When you enter a chat in Honk, you’ll be presented with two large conversation bubbles. The gray one on the top is where your friend’s messages are shown, while you type in the blue one. (You can change the colors and theme, if you choose.)

As you type, the other person will see the text you’re entering into this box in real-time — including the pauses and typos that would normally be missed. This “live typing” experience is reminiscent of older communication technology, like the early instant messaging app ICQ, or the innovative collaboration tool Google Wave, for example.

In Honk, you’re given 160 characters to type out your thoughts, and this is counted down on the right side of screen below the conversation bubbles. But you don’t tap a “Send” button to share the message — the recipient saw the text as it was entered, after all. Instead, you just tap the double arrow “refresh” button to clear the screen and type something new.

There are also buttons for sending emoji, snapping a photo or accessing photos from your Camera Roll to share in the chat. The emoji here work more like iMessage’s “Send with Echo” screen effect, as you’re not just sending a single emoji when you use this feature — you’re sending several huge emoji that temporarily fill the screen.

You can also optionally assign emoji to any word or phrase within an individual chat, using a “Magic Words” feature that will trigger effects as you type (see above). Plus, you can customize chat themes on a per-conversation basis or turn off notifications from an individual user, if you don’t want to hear from them as much.

None of the conversations are stored and there’s no history to look back on. This is similar to messaging apps like Snapchat or Messenger’s Vanish Mode, for instance. (Honk hasn’t clarified its position on security, however, so proceed with caution before getting into riskier content.)

And, of course, if you need to get someone’s attention, you can tap “Honk” to flood them with notifications.

If this all seems somewhat silly, then you’re probably not the target market for the Honk messaging experience.

The app is clearly aiming for a young crowd of largely teenage users. When Honk asks for your age during setup, in fact, you can select an exact number from the list that appears — unless you’re “old,” that is. The last option on the list of ages is “21+” — the “older folks” age bracket, which may sting a bit for the millennial crowd who often still think of themselves as the online trendsetters.

But Honk is aiming to grab Gen Z’s interest, it seems. It’s even marketing to them on TikTok, where it’s already generated some 140K+ “Likes,” as of the time of writing, despite having only uploaded its first video yesterday. Honk founder Benji Taylor also noted on Twitter the app has seen 550,000 “Honks” sent so far, as of Wednesday, December 23, 2020, shortly after noon Eastern.

@usehonkwait for it ##fyp♬ original sound – Honk

Per its website, Honk is the flagship product from software company and app publisher Los Feliz Engineering (LFE), which is backed by investors including Naval Ravikant, Elad Gil, Brian Norgard, David Tisch, Jeff Fagnan, Ryan Hoover, Sarah Downey, Josh Hannah, Sahil Lavingia and others.

“It’s exceptionally well designed,” said Product Hunt founder and Weekend Fund investor Ryan Hoover, about Honk. “[Honk founder] Benji [Taylor] and team labored over the small details, from the animations to the sounds. They’re also super focused on speed,” he added.

Taylor declined a full interview when TechCrunch reached out, noting the team was focused on building the product for the time being.

“We’ve been working on Honk for a while now. Our goal is to make messaging fun, and empower people to communicate in new, creative ways that take relationships deeper,” Taylor told TechCrunch. “Ultimately though, we’re a small team building this for ourselves and our friends. If other people like it, all the better,” he said.

Honk, we should note, has been struggling under the load of new signups at launch and high usage. Honk users report the app will sometimes say they’re offline when they’re not, for example, among other bugs. Honk acknowledged the issues on its Twitter and says it’s been working to resolve them.

The app is currently a free download on iOS. It does not include in-app purchases or have any obvious business model.


Source: Tech Crunch

Letterhead wants to be the Shopify of email newsletters

You’re probably investing in an email newsletter these days, whether you’re an international brand, a nonprofit, or a local news publisher.

Maybe email is even your focus now, because you got burned by Facebook, Google or other closed platforms during the past decade. The problem is that the tools you have available are probably too generic, or are built specifically for marketers. What if you want to make money from the newsletter content itself?

Letterhead is slicing through the vast market of existing email SaaS products, betting that a cross-section of revenue and collaboration needs are not being met properly for newsletter creators of all types. Instead, it puts all ad sales, paid subscriptions and newsletter content management into a single, streamlined product.

Its viewpoint on the future of newsletters — and its customer base so far — are intriguingly different from your typical SaaS startup in Silicon Valley. And there’s a reason for that. Letterhead is actually a product spinout of a community publisher in Miami called WhereBy.Us that began life in 2014 by launching a local media site, The New Tropic. The publication became a rare success in online local news, once it focused on the email newsletter format.

“Initially, our goal was to create a local media product that would help people learn about the city, get more involved and serve a new generation of local news users,” co-founder and CEO Chris Sopher tells me by video chat. “Our ideas had included opening a bar, events, and all kinds of other stuff. We quickly pared that back to the things that were working, and email was at the top of the list.”

Advertising inventory in these emails was in high demand, so the company built out a self-service payment system for advertisers, that allowed its newsletter writers to easily publish the correct ads in the correct place.

With this business model and technology as a foundation, it launched or acquired newsletters in Seattle, Portland, Orlando and Pittsburgh. Through this process, it has continued to improve the tool itself.

It also discovered the broader demand.

“People would reach out to us and say ‘I love this newsletter, what tools do you use?’ because it was such a pain to produce emails with all of the different tools out there,” Sopher explains.

(Those tools, in this author’s experience, generally include a combination of Mailchimp, Constant Contact or Sailthru, together with your main web publishing CMS like WordPress, your separate subscription software like Piano and however you are managing ads.)

“We’d tell people that we used our own internal tools and they’d say ‘oh, can we use those, too? And we’d say ‘no, that’s not what we’re doing.’ Eventually, we said no enough that we looked at each other and said, ‘we should figure out how to get to yes on this.’ And that’s where Letterhead came from.”

Today, Whereby.us is one of the few success stories from a catastrophic decade in local news. Sopher says that its five city newsletters are comfortably profitable via ads overall — having recovered from a pandemic dip earlier this year — and are continuing to grow.

But the new focus is on Letterhead’s tools, including the ad system, a new paid subscription feature that lets you do things like add paywalled subsections of emails, easy-to-use text editing and template formatting, and soon, analytics.

“Sponsorships and ads were [needs] we heard about most, so that’s where we started,” cofounder and COO Rebekah Monson said by email. “The bigger vision is to create a set of tools and services that feed into each other in one easy place, and help all of those revenue streams grow, eventually branching out from email. We’re seeing demand for that not just from traditional media publishers, but also from marketers, nonprofits, universities, professional associations — all these folks who have engaged communities and want to deepen those relationships and bring in revenue through that engagement.”

On the spectrum of email newsletter products, Letterhead’s focus on revenues and team collaboration places it adjacent to Substack’s focus on the individual writer, and to other products like Lede designed specifically around subscription news organizations.

Letterhead is explicitly a hosted software solution that you pay for your organization to use, not an open-source project like Ghost. Like how Shopify provides a suite of white-label ecommerce features for anyone who wants to run an online business, it wants to be the engine humming away under the hood of your newsletter.

On the much broader spectrum of all email solutions in the SaaS world, Letterhead is betting that its understanding of the market and its product design can beat out the brutally competitive world of SaaS email products.

Since soft-launching earlier this year, Sopher says it has already been signing up a broad range of customers. Examples he cited include startups (Shoot My Travel), nonprofits (Vida y Salud and Refresh) and political groups (OD Action) as customers, as well as local news publishers, of course (VTDigger, Choose954, and Santa Cruz Local). Letterhead is also a partner in WordPress’ News Pack program, which is a collection of plugins for publishers on that CMS.

“We’re always going to have an affinity to media publishing… but there is a broader need than just that industry,” he explains. “And we’re also seeing this moment where a lot of other organizations are participating in [publishing]. You name a topic, there will be professional reporters out there doing great work. But it’s also pretty likely that there is a brand, an agency, a nonprofit, or some other organization creating interesting and useful content, and building a community around it — that would not raise its hand and say that it is part of the news industry.”

Sopher also notes that the product is designed to be modular, so that companies can just use parts of it and integrate its features with other email service providers and most any tech stack.

“What we’re seeing is that smaller customers are coming on for the simplicity of having it all in one place without sacrificing monetization,” he adds, “but larger customers are choosing us as one part of their stack as they build a multifaceted business or grow out of tools geared more to individual or independent creators.”

With the revenue options in place, Sopher says analytics and additional ESP options are coming next.

Over the course of its history, Whereby.us has raised $5 million from across tech and media. Backers include the Knight Foundation, Jason Calacanis’ LAUNCH fund, Band of Angels, McClatchy, hundreds of smaller investors via Republic and SeedInvest, and most recently a round led by Brick Capital.


Source: Tech Crunch

Dear Sophie: What’s ahead for US immigration in 2021?

Here’s another edition of “Dear Sophie,” the advice column that answers immigration-related questions about working at technology companies.

“Your questions are vital to the spread of knowledge that allows people all over the world to rise above borders and pursue their dreams,” says Sophie Alcorn, a Silicon Valley immigration attorney. “Whether you’re in people ops, a founder or seeking a job in Silicon Valley, I would love to answer your questions in my next column.”

Extra Crunch members receive access to weekly “Dear Sophie” columns; use promo code ALCORN to purchase a one- or two-year subscription for 50% off.


Dear Sophie:

I’m in people ops and our team is trying to plan ahead for immigration in the new year and beyond.

What’s ahead for U.S. visas and green cards?

—Ready in Redwood City

Dear Ready:

Ha! I love it. Well, although I don’t have a crystal ball (yet), there’s a lot of opportunity, predictability and security that we can anticipate for immigration ahead.

Our U.S. immigration policy will experience a tremendous growth spurt in the coming months as Trump completes his regulatory agenda, litigation culminates and Biden takes office on January 20. The changes I’m tracking will incentivize U.S. companies to hire and retain top global talent and will make it easier for them to do so. There are also going to be increased opportunities for families and founders, strengthening the U.S. and Silicon Valley tech startup communities.

We can anticipate that the first 100 days of President-elect Biden’s term will focus on undoing many Trump-era immigration changes. Some of this will happen by executive order (although probably not tweets!) and some of it will be required to follow the procedures set forth in law through the Administrative Procedure Act (APA). The APA governs the process by which federal agencies develop and issue regulations.

Following procedures to rescind or amend rules already put into place — even on an expedited basis — takes time to allow for adequate review and public comments. We can anticipate that due process will unfold to effectuate these changes.


Source: Tech Crunch

No rules, no problem: DeepMind’s MuZero masters games while learning how to play them

DeepMind has made it a mission to show that not only can an AI truly become proficient at a game, it can do so without even being told the rules. Its newest AI agent, called MuZero, accomplishes this not just with visually simple games with complex strategies, like Go, Chess, and Shogi, but with visually complex Atari games.

The success of DeepMind’s earlier AIs was at least partly due to a very efficient navigation of the immense decision trees that represent the possible actions in a game. In Go or Chess these trees are governed by very specific rules, like where pieces can move, what happens when this piece does that, and so on.

The AI that beat world champions at Go, AlphaGo, knew these rules and kept them in mind (or perhaps in RAM) while studying games between and against human players, forming a set of best practices and strategies. The sequel, AlphaGo Zero, did this without human data, playing only against itself. AlphaZero did the same with Go, Chess, and Shogi in 2018, creating a single AI model that could play all these games proficiently.

But in all these cases the AI was presented with a set of immutable, known rules for the games, creating a framework around which it could build its strategies. Think about it: if you’re told a pawn can become a queen, you plan for it from the beginning, but if you have to find out, you may develop entirely different strategies.

This helpful diagram shows what different models have achieved with different starting knowledge.

As the company explains in a blog post about their new research, if AIs are told the rules ahead of time, “this makes it difficult to apply them to messy real world problems which are typically complex and hard to distill into simple rules.”

The company’s latest advance, then, is MuZero, which plays not only the aforementioned games but a variety of Atari games, and it does so without being provided with a rulebook at all. The final model learned to play all of these games not just from experimenting on its own (no human data) but without being told even the most basic rules.

Instead of using the rules to find the best-case scenario (because it can’t), MuZero learns to take into account every aspect of the game environment, observing for itself whether it’s important or not. Over millions of games it learns not just the rules, but the general value of a position, general policies for getting ahead, and a way of evaluating its own actions in hindsight.

This latter ability helps it learn from its own mistakes, rewinding and redoing games to try different approaches that further hone the position and policy values.

You may remember Agent57, another DeepMind creation that excelled at a set of 57 Atari games. MuZero takes the best of that AI and combines it with the best of AlphaZero. MuZero differs from the former in that it does not model the entire game environment, but focuses on the parts that affect its decision-making, and from the latter in that it bases its model of the rules purely on its own experimentation and firsthand knowledge.

Understanding the game world lets MuZero effectively plan its actions even when the game world is, like many Atari games, partly randomized and visually complex. That pushes it closer to an AI that can safely and intelligently interact with the real world, learning to understand the world around it without the need to be told every detail (though it’s likely that a few, like “don’t crush humans,” will be etched in stone). As one of the researchers told the BBC, the team is already experimenting with seeing how MuZero could improve video compression — obviously a very different problem than Ms. Pac-Man.

The details of MuZero were published today in the journal Nature.


Source: Tech Crunch

Fluent Forever raises $4.9M for its language learning system

Fluent Forever, a startup that uses a novel learning system to help its users master a new language faster, has raised a $4.9 million funding round led by Denver-based Stout Street Capital. Other investors in this round include The Syndicate, LAUNCH, Mana Ventures, Noveus VC, Flight.VC, Insta VC, UpVentures, Firebrand Ventures, Cultivation Capital, Spero Ventures and Lofty Ventures.

In many ways, Fluent Forever is a direct competitor to Duolingo, Babbel and similar online language learning services. What sets it apart is a focus on a personalized learning system that emphasizes ear training, visual aids and something akin to spaced-repetition for helping you memorize new words and phrases. It’s a paid service (after a 14-day free trial) with subscriptions starting at $10 per month for a monthly subscription and the usual discounts for longer-term commitments.

To teach himself his first languages, the company’s founder and CEO Gabriel Wyner used the popular flashcard service Anki, wrote a book about his approach, and taught workshops on language learning using his system with Anki. But as he noted, Anki is a serious tool, and simply learning how to get the most out of it takes a lot of time and energy.

Image Credits: Fluent Forever

“I’ve watched everyone else fail at language learning,” he told me. “And the first thought is, okay, well, if you just learn how to do it right, then that’s a fixable thing. That’s exciting. And then once you have a solution for people and they’re all excited about it — but then you watch them fail because of IT reasons. That’s extra frustrating.”

In many ways then, Fluent Forever uses Wyner’s flashcard approach — because building those flashcards by hand is at the core of his learning system — and turns it into a far-easier-to-use application.

What people want, Wyner acknowledged, is a tool where you just press some buttons and learn something. But that doesn’t work. “I had to have a really strong reaction to this — a really strong answer — and say, ‘absolutely not. That is the one thing that teaches you is building it.’ ”

Wyner is not afraid to compare his approach to Duolingo’s and argues that its focus on translation exercises doesn’t translate to real language skills in the long run. At the same time, he freely acknowledges that the Duolingo user experience and gamification are far better than Fluent Forever. But he also believes that learners see far better results with his system.

Image Credits: Fluent Forever

“We ask [our users]: ‘Why are you with us? Why would you pay for us when you could just get Duolingo for free?” What they come back with is, ‘yeah, your product is rough around the edges. I wish you would fix this, this and that, but you had me thinking in Spanish in two weeks,” Wyner said.

Fluent Forever currently supports nine languages: Japanese, French, Russian, Mexican and Spanish Spanish, Italian, Korean, German and Brazilian Portuguese, with Dutch being the next language the team is tackling.

As Wyner told me, the company had trouble raising in 2019, in part because the service was seeing pretty flat growth at the time. “People are very skeptical about language learning — that is not a sexy field. People don’t like it. The idea of jumping and trying to be competitive with Duolingo was just not appealing to anyone,” he told me. Come 2020, though, growth picked up, even before the COVID pandemic. At the same time, Fluent Forever also participated in Jason Calacanis’ Launch Accelerator.

Looking ahead, Wyner tells me that Fluent Forever is looking at ways to bring live tutors into the loop. Live tutoring online has been done before, of course, and there are some companies like Preply that specialize in it already, but what Fluent Forever wants to do is combine the online language learning service with short live sessions and then use the online component to go back to that conversation over the course of a week or so. One advantage here is that these users — who will likely pay a premium for the live service — will also use their time with live tutors to create their own personalized sentences in the Fluent Forever system, which could then over time become content that’s available to all users, too.


Source: Tech Crunch