Editor’s note: Want this in your inbox every Saturday morning? Click here to subscribe to Startups Weekly and all the other great TechCrunch newsletters.

Remember when the top investors and companies in tech were reacting to SoftBank’s every move? These days, we are picking through the latest results from the Japanese conglomerate and its Vision Fund to see how things went wrong, and where it is still succeeding with its startup portfolio.

First up, this fund appears to be out of additional money to spend, as Arman Tabatabai and Danny Crichton found buried in the footnotes of its new regulatory filing. Meanwhile, as they tallied on TechCrunch, the losses have piled up lately:

The Vision Fund officially lost $17.4 billion in value according to SoftBank’s financials for the year ending this past March 31. The year before, SoftBank had registered a positive gain in the Vision Fund’s value of $12.8 billion, which means that the damage of this year’s performance has completely wiped out all gains the fund had made in the previous year. But the real shock is the performance of the fund’s underlying portfolio companies. The Vision Fund currently has 88 active portfolio companies that have not exited. Of those, 19 investments saw a gain in combined value of $3.4 billion according to SoftBank, while 50 companies saw a decline in value aggregating to $20.7 billion in losses. 19 portfolio companies were left unchanged in value.

Is this worse than what the rest of the market at large is going through? Here’s Alex Wilhelm’s view on Extra Crunch:

To some degree this feels counter-narrative. Tech shares have rebounded in recent weeks, rebuilding sentiment in the sector — perhaps the COVID-19 downturn won’t be that bad, the thinking seems to go. The SoftBank Vision Fund’s results paint a more negative picture of the economy: It’s bad in many areas, lots of companies are impacted and the value of many unicorns is too high, even if the scale of write-downs that private investors like venture capitalists will have to endure is not yet clear. The private market can, therefore, expect a host of down-rounds if unicorns need to raise capital in the short-term. And many will. The Vision Fund report card, then, is an indication that enterprise software is doing as well as we might have thought, that there are some winners in the health-tech space and that, aside from those exceptions, the rule appears to be a downturn in startup land. 

Emphasis mine. Arman and Danny also broke out Arm’s financials for EC and what that top SoftBank company shows about the future of semiconductors. And, for both education and amusement, they provided a commentary about SoftBank’s in-depth and sometimes bizarre presentation about the results.

The symbolism of Jack Ma’s SoftBank board resignation

Masayoshi Son made his name via a seminal bet on a very young Alibaba back in the 1990s, and since then he and SoftBank have had much of their net value and stature tied up in the success of Jack Ma’s efforts. Ma, in turn, has bolstered SoftBank by holding a board seat on the conglomerate since 2008. After 14 years and broadly changing interests on both sides, it’s not surprising that he resigned. But as Danny wrote for TechCrunch in a helpful sidebar to the other Softbank coverage:

[I]t’s not just about an investor and his entrepreneur breaking some ties after two decades in business together. It’s about the fraying of the very globalization that powered the first wave of tech companies — that a Japanese conglomerate with major interests in the U.S. and Europe could invest in a Hong Kong/China startup and reap huge rewards. That tech world and the divide of the internet and the world’s markets continues unabated.

What will save college-town startup hubs?

Few people alive remember, but Palo Alto used to be considered a long way from San Francisco… this was back when Stanford University actually was a farm, though. The interplay of the university’s technical research and education with local technologists was core to how Silicon Valley formed and how the region grew, and in recent decades many other metros of all sizes have implemented their own successful versions of this playbook.

But maybe pandemic effects will cause startup activity to contract to the biggest startup hubs? In this week’s staff survey (a new format we’re trying out), Danny believes that’s the case. The revenues for universities will be hit too hard by the loss of foreign student tuition, decreased attendance domestically due to closed campuses and student financial problems, etc. Natasha Mascarenhas looks back at her own experiences and finds the in-person experience so irreplaceable that she thinks the core attendance will recover. Alex agrees with that.

As I move out of the Bay Area to a college town this weekend, I think I disagree with all of the above. Yes, I also expect higher education to get slammed — but what is going to remain? STEM programs already have government and private funding lined up that can stretch many years into the future, and these schools have wealthy, supportive alumni and can generate revenue from commercialization (aka startup creation). Which means that, as much as anything will exist anywhere physically in higher ed, the research labs and science and engineering programs of the country (and the world) will continue to operate. The tech companies that are still booming publicly or privately will need to hire more graduates with these degrees. So, even with remote learning, the core institutions and their environs will have the means to continue, and be regular destinations for tech talent.

Danny, it is the big cities that I think will get slammed the hardest, especially those with troubled local and state revenue sources like here in California. People of all income levels were already fleeing the largest metros due to high prices, now the pandemic is reinforcing that they can work remotely with little to no drop in productivity. Instead, commercial real estate, typically a key urban tax base, is in free-fall. Let’s say you work in tech but want to spend less and have more space and amenities. Yes there are many suburbs and exurbs you can move to — but the college town ones are some of the nicest. Nobody is fleeing Boulder now. But I bet a lot of people wish they could move there.

Combine all of this with the global networking tools that the tech industry has been hard at putting together, and I think finding a cofounder and building a company will soon be as easy as finding an online date. Why not find yourself a nice garage in a sleepy college town like Bill Hewlett and David Packard did not so long ago and settle in for some hardcore entrepreneurship? Find your cofounders and key employees from near and far as you please, and enjoy the benefits of your alma mater’s local network. Just make sure you have a great wifi connection and an ergonomic workstation.

Delivery robot demand starts to grow, create human jobs

Automation turns out to still require a lot of blood, sweat and tears to operate correctly. Resident automotive expert Kirsten Korosec takes a look at how the delivery robot sub-sector of autonomous vehicles has been hiring remote humans to help delivery robots navigate the trickiest parts of a route safely as demand grows during the pandemic. Her main example in this in-depth look on TechCrunch is a partnership between Postmates and a startup called Phantom Auto, which focuses on AV teleoperations.

Using Phantom Auto’s software, a Postmates fleet supervisor can monitor a robot from thousands of miles away. The supervisor will jump in to help the bot navigate the first and last 15 feet to a restaurant or the recipient or if it needs help crossing a busy street.

These robot guides can assist using a couple of methods. The human teleoperator can provide input to the system, something as simple as a thumbs up or thumbs down to help the bot make the right choice. The employee can also use a hand-held remote controller to steer, accelerate or slow down the bot in real-time.

The teleoperations component of mobility is spreading more broadly. She separately covered a scooter company in Atlanta that is hiring remote operators in Mexico City to deliver the vehicles to customers.

If you’re focused on these topics, you might be interested in the other things Kirsten is up to as well (if you’re not reading her already). In addition to her regular coverage, she’s been doing surveys of mobility investors along with Megan Rose Dickey for Extra Crunch. We published the first last week on the larger impact of the pandemic on the sector. Kirsten also has a weekly free newsletter called The Station about the topic and her coverage, which you can read and subscribe to here.

Investors surveyed on enterprise software, cannabis

Pandemic or no, enterprise investors will not stop being bullish, thank you very much. Resident enterprise reporter Ron Miller caught up with top investors in the space in the space for the first of a series on the cloud that he has coming. Here’s a money quote from the Extra Crunch article, courtesy of Max Gazor at CRV.

It’s abundantly clear that cloud software markets are bigger than most people anticipated. We continue to invest heavily there as we have been doing for the last decade. Specifically, the most exciting trend right now in enterprise is low-code software development. I’m on the board of Airtable, where I led the Series A and co-led the Series B investments, so I see first-hand how this will play out. We are heading toward a future where hundreds of millions of people will be empowered to compose software that fits their own needs. Imagine the productivity and transformation that will unlock in the world! It may be one of the largest market opportunities we have seen since cloud computing.

And now for something completely different. Cannabis has emerged as a serious half-legal sector that few of us have qualms about, in this part of the world at least. It has tended to breed its own strain of investor — many of whom Matt Burns caught up with for our second survey this week. The pandemic seems to have turned things around for the category, at least according to some. Here’s Matt Hawkins of Entourage Effect Capital:

Cannabis went from illegal to essential in about two weeks flat — cannabis is now listed right alongside hospitals, doctors, grocery stores, gas stations and fire departments as an essential service. As we edge close to federal legalization, there is still a large demand for research on cannabis’ medicinal benefits and a lot more opportunities to create cannabis-derived medicines. There is a lot to be excited about in the long term.

Across the week

Extra Crunch

What to do when your VC writes your startup off

Why VCs say they’re open for business, even if they’re pausing new deals

GitLab’s head of Remote on hiring, onboarding and why Slack is a no-work zone (part 1)

GitLab’s head of Remote on what people tend to get wrong about remote work (part 2)

Popping the hood on Vroom’s IPO filing

The Great Reset

TechCrunch

Work From Home is dead, long live Work From Anywhere

Following Luckin Coffee scandal, Nasdaq ready to tighten rules on IPO listings

How I Podcast: Articles of Interest’s Avery Trufelman

Europe to Facebook: Pay taxes and respect our values — or we’ll regulate

How to decode a data breach notice

Around TechCrunch

TechCrunch Disrupt 2020 is going virtual

Startup Battlefield is going virtual with TechCrunch Disrupt 2020

Sequoia Capital’s Roelof Botha is coming to Disrupt this fall

Extra Crunch Live: Join Verizon CEO Hans Vestberg for a live Q&A May 26 at 2pm ET/11am PT

Extra Crunch Live: Join Box CEO Aaron Levie May 28th at noon PT/3 pm ET/7 pm GMT

#EquityPod: Clubhouse proves that time is a flat circle

Listen here.

From Alex:

Hello and welcome back to Equity, TechCrunch’s venture capital-focused podcast, where we unpack the numbers behind the headlines.

First, a big thanks to everyone who took part in the Equity survey, we really appreciated your notes and thoughts. The crew is chewing over what you said, and we’ll roll up the best feedback into show tweaks in the future.

Today, though, we’ve got Danny and Natasha and Chris and Alex back again for our regular news dive. This week we had to leave the Vroom IPO filing, Danny’s group project on The Future of Work and a handwashing startup (?) from Natasha to get to the very biggest stories:

• Brex’s $150 million raise: Natasha covered the latest huge round from corporate charge-card behemoth Brex. The party’s over in Silicon Valley for a little while, so Brex is turning down your favorite startup’s credit limit while it stacks cash for the downturn.
• Spruce raises a $29 million Series B: Led by Scale Venture Partners, Spruce is taking on the world of real estate transactions with digital tooling and an API. As Danny notes, it’s a huge market and one that could find a boost from the pandemic.
• MasterClass raises $100 million: Somewhere between education and entertainment, MasterClass has found its niche. The startup’s $180 yearly subscription product appears to be performing well, given that the company just stacked nine-figures into its checking account. What’s it worth? The company would only tell Natasha that it was more than $800 million.
• Clubhouse does, well, you know. Clubhouse happened. So we talked about it.
• SoftBank dropped its earnings lately, which gave Danny time to break out his pocket calculator and figure out how much money it spent daily, and Alex time to parse the comedy that its slideshow entailed. Here’s our favorites from the mix. (Source materials are here.)

And at the end, we got Danny to explain what the flying frack is going on over at Luckin. It’s somewhere between tragedy and farce, we reckon. That’s it for today, more Tuesday after the holiday!

Source: Tech Crunch

The global pandemic has cast a light on decades of cumulative efforts to manipulate and suppress voters, showing that the country is completely unprepared for any serious challenge to its elections system. There can be no more excuses: Every state must implement voting by mail in 2020 or be prepared to admit it is deliberately sabotaging its own elections. (And for once, tech might be able to help.)

To visualize how serious this problem is, one has only to imagine what would happen if quarantine measures like this spring’s were to happen in the fall — and considering experts predict a second wave in that period, this is very much a possibility.

If lockdown measures were being intensified and extended not on May 3rd, but November 3rd, how would the election proceed?

The answer is: it wouldn’t.

There would be no real election because so few people in the country would be able to legally and safely vote. This is hardly speculative: We have seen it happen in states where, for lack of any other option, people had to risk their lives, breaking quarantine to vote in person. Naturally it was the most vulnerable groups — people of color, immigrants, the poor and so on — who were most affected. The absurdity of a state requiring voters to gather in large groups while forbidding people to gather in large groups is palpable.

With this problem scaled to national levels, the entire electoral process would be derailed, and the ensuing chaos would be taken advantage of by all and sundry for their own purposes — something we see happening in practically every election.

For the 2020 election, if any elections official in this country claims to value the voters for which they are responsible, voting by mail is the only way to enable every citizen to register and vote securely and remotely. Anything less can only be considered deliberate obstruction, or at best willful negligence, of the electoral process.

There’s a fair amount of talk about apps, online portals and other avenues, and these may figure later, but mail is the only method guaranteed right now to securely serve every address and person, providing the fundamental fabric of connectivity that is absolutely necessary to universally accessible voting.

Hand-wringing about fraud, lost ballots and other issues with voting by mail is deliberate, politically motivated FUD (and you can expect a lot of it over the next few months). States where voting by mail is the standard report no such issues; on the contrary, they have high turnout and few problems because it is simple, effective and secure. As far as risk is concerned, there is absolutely no comparison to the widespread and well-documented process and security issues with touchscreen voting systems, even before you bring in the enormous public health concerns of using those methods during a pandemic.

Federal law requires that troops around the world, among others unable to vote in person, are able to request and submit their ballots by mail. That this is the preferred method for voting in combat zones is practically all the endorsement such a system needs. That the president votes by mail is just the cherry on top.

Fear of voters

So why hasn’t voting by mail been adopted more widely? The same reason we have gerrymandered districts: Politicians have manipulated the electoral process for decades in order to stack the deck in their favor. While gerrymandering has been employed with great (and deplorable) effect by both Democratic and Republican officials, voter suppression is employed overwhelmingly by the political right.

While this is certainly a politically charged statement, it’s not really a matter of opinion. The demographics of the voting public are such that as the proportion of the population that votes grows, the aggregate position begins to lean leftward. This happens for a variety of reasons, but the result is that limiting who votes benefits conservatives more than liberals. (I am not so naive to think that if it were the other way around, Democrats would altogether abstain from the practice, but that isn’t the case.)

This is not a new complaint. Deliberate voter suppression goes back a century and more. Nor is the practice equally distributed. For one thing, white, well-off, urban areas are more likely to have effective and modern voting systems and laws.

This is not only because those areas are generally the first to receive all good things, but because voter suppression has been aimed specifically at people of color, immigrants, the poor and so on. Again, this is no longer a controversial or even particularly partisan statement; it has been admitted to by politicians and strategists at every level — including, quite recently, by the president: “They had things, levels of voting that if you’d ever agreed to it, you’d never have a Republican elected in this country again.”

When voting by mail was merely a convenient, effective alternative to voting in person, it was fairly easy to speak against it. Now, however, voting by mail is increasingly looking like the only possible method to accomplish an election.

Again, think of how we would vote during a stay-at-home order. Using only today’s methods would be dangerous, chaotic and generally an ineffective way to ask the population at large who they want to lead their city, state and country.

That is no way to conduct an election. Therefore, we currently have no way to conduct a national election. Voting by mail is the only method that can realistically be rolled out to accomplish an effective election in 2020.

Disunited states

Because elections are run by state authorities, voting methods and laws vary widely between them. The quickest way to a nationwide vote-by-mail system would use federal funding and authority, but even if states were in favor of this (they won’t be, as it is an encroachment on their authority), Washington is not. The possibility of a bill implementing universal voting by mail passing the House, Senate and the president’s desk by November is, sadly, remote.

Which is not to say that no one in D.C. is not trying it:

This means it’s down to the states — not great news, considering it is at the state level that voting rights have been eroded and voter suppression enshrined in policy.

The only hope we have is for state authorities to recognize that the 2020 presidential election will be a closely watched litmus test for competence and corruption that will haunt them for years. It’s one thing to put your finger on the scale under normal circumstances. It’s quite another to author a high-profile electoral failure in an election few doubt will be one of the most consequential in American history — especially if that failure was manifestly preventable.

And we know it is preventable because due to federal voting rights laws, every state already has some form of accessible, mail-in or absentee voting. This is not a matter of inventing a new system from scratch, but scaling existing, proven systems in ways already demonstrated and verified over decades. Several states, for instance, have simply announced that all voters will get absentee ballots or applications sent unrequested to their homes. No one said it would be easy, but the first step — committing — is at least simple.

It will be obvious in a few months which state authorities actually care about the vote and which see it as just another instrument to manipulate in order to retain and accrue power. The actions taken in the run-up to this election will be remembered for a long time. As for the federal government interfering with states’ prerogative to run their own elections — that’s a violation of states’ rights that I expect will encounter strong bipartisan opposition.

How tech can help without hindering

The tech world will want to aid in this cause out of several motives, but the simple truth is there’s no way a technological solution can be developed and deployed by November. And not only is it infeasible, but there is serious political opposition to online voting systems to be widely deployed. The idea is a non-starter for this election and probably the next.

Rather than trying, Monolith-style, to evolve voting to the next phase by taking on the whole thing tip to tail, tech should be providing support structures via uniquely digital tools that complement rather than replace effective voting systems.

For example, there is the possibility, however remote, that a mailed ballot will be intercepted by some adversary and modified, shredded, selectively deposited, or what have you. No large-scale fraud has ever been perpetrated, despite what opponents of voting by mail might say. States developed preventative solutions long ago, like secure ballot boxes placed around the city and tamper-evident envelopes.

But end to end security is something at which the tech sector excels, and moreover recent advances make a digitally augmented voting process achievable. And there’s plenty of room for competition and commercial involvement, which sweetens the pot.

Here’s a way that commonplace tech could be deployed to make voting by mail even more secure and convenient.

Imagine a mail-in ballot of the ordinary fill-in-the-bubble type. Once a person makes their selections, they take a picture of the ballot in a dedicated, completely offline app. Via fairly elementary image analysis nearly any phone can now perform, the votes can be detected and tabulated, verified by the voter, then hashed with a unique voter sheet ID into a code short enough to be written down.

The ballot is mailed and (let us say for now) received. When it is processed, the same hash is calculated by the machine reader and placed on an easily accessible list. A voter can check that their vote was tabulated and correctly recorded by entering their hash into a website — which itself reveals nothing about their vote or identity.

What if something goes wrong? Say the ballot is lost. In that case the voter has a record of their vote in both image and physical form (mail-in ballots have little tear-off tabs you keep) and can pursue this issue. The same database that lets them verify their vote was correct will allow them to see if their vote was never cast. If it was interfered with or damaged and the selections differ from what the voter already verified, the hash will differ, and the voter can prove this with the evidence they have — again, entirely offline and with no private information exposed.

This example system only works because smartphones are now so common, and because it is now trivial to process an image quickly and accurately offline. But importantly, the digital aspect only addresses shortcomings of the mail-in system rather than being central to it. You vote with only a ballpoint pen, as simply as possible — but if you want to be sure, you may choose to employ the latest technology to track your vote.

A system like this may not make it in time for the 2020 election, but voting by mail can and must if there is to be an election at all.

Source: Tech Crunch

Working from home isn’t going anywhere anytime soon, and a slew of companies just announced longer-term initiatives to make their remote work practices either extend or permanent. That means for some it’s the perfect time to take their at-home videoconferencing setup even further, so we’re going to take a closer look at various core elements to build on our initial exploration of what can help you improve your video call or live broadcasting game. Today, it’s all about audio.

Microphone basics

In our initial feature, I highlighted some great entry-level options for add-on pics that you can use to produce better sound than what your Mac or PC can produce alone. Those included the Samson Meteor Mic, a longstanding favorite that connects directly via USB and that produces great, full-bodied sound without any customization required.

There’s also the Rode Wireless GO, a simple and affordable wireless mic pack kit that you can use on its own, or pair with a lavalier like the Rode Lavalier GO for a bit better sound. Rode also makes a great USB mic, that, like the Meteor Mic, just works and comes in at around $100 – the Rode NT-USB Mini. It features some design decisions like a magnetic desk stand that could make it more flexible for use for certain setups vs. the Meteor Mic, and the sound it produces is also fantastic.

To improve your Rode Wireless GO setup a bit further, or to use a wired lavalier-style wearable mic plugged directly into your computer or audio interface, there are a couple of great options available from Sennheiser that provide subtle but noticeable sound quality improvements no matter how you’re using them.

The Sennheiser MKE Essential Omni is a great lavalier mic that’s often used in stage productions and other professional settings, with a tiny profile that you can pretty easily hide in clothing using the included clip, or even in hair, or in tandem with an earset holder for putting it right on your cheek next to your mouth. You’ll get slightly different sound profiles depending on how you wear it, but it generally produces great, warm sound and doesn’t cost too much at just under $200 (on the relative scale of sound equipment prices).

Sennheiser’s ME 2-II is another, lower-cost option at $129.95 that also produces great results, and works with with wireless transmitters like the Rode Wireless GO, but it’s a bit less warm and present than the MKE Essential.

Getting serious about sound

High-end lavalier mics are already starting to get into high expense territory, but as with most audio equipment, the sky’s the limit here. That’s also true for shotgun microphones, which is another option for rigging your setup for the best possible audio without compromising on things like unsightly microphones in frame, or some of the trade-offs that come with using very physically small microphones like lavalier and lapel mics.

In our original post, I talked about using a Rode VideoMic NTG as one option, and that is indeed a great, mid-level shotgun mic to experiment with, with the added benefit of being terrific for use on-camera in the field thanks to its built in battery, compact dimensions and intelligent compatibility with a range of modern cameras.

But for home studio use, there are shotgun mics that are much more appropriate to the task. The Rode NTG3 is a personal favorite, and a popular standard in the broadcast and film industries – for good reason. The NTG3 is a tubular mic with a standard XLR output, that required 48v phantom power and that is perfect for videos shooting scenarios where you’re staying relatively still in a fixed location with cameras also mounted in fixed positions – ie., exactly how most people have their home working spaces set up.

The Rode NTG3 is a bit of a budget-buster, however – it’s $699, which is more than even some very high-quality standard podcasting mics out there. But for the price, you get an extremely high-quality piece of hardware, that has built-in moisture resistance for shooting outdoors if that’s ever something you want to do, and that sounds great even when mounted out of sight beyond the frame of your camera’s lens.

It’s also supercardioid in its pickup pattern, which means it does an excellent job of picking up sound directly in front of it, but not sound to either side. That’s a great advantage to have in most shared home office spaces, just like it is with on-location film shoots.

Another top option that’s a popular favorite, and that comes in at a lower price point, is the Sennheiser MKE 600. At around $330, it’s roughly have the price of the NTG3, and it has a built-in battery in case you want to take it with you and plug it into your camera. It also uses XLR, which means you’ll need a preamp like the Focusrite 2i2 or the recently released Audient EVO 4 to make it work with your computer (or the iRig Pre if you’re running it to a deck like the Blackmagic ATEM Mini, as I was).

The sound from the MKE 600 is still top notch, but it doesn’t do quite as good a job as the NTG3 of eliminating any self-noise, and of capturing a deep, rich tone that’s suitable to deeper voices. You can check out a comparison of both boom mics, along with the Sennheiser MKE Essential, in the video below.

Another option is to use a pole or boom-mounted mic like you generally see podcasters or radio personalities use. These include popular options like the Shure SM7B, which you’ll probably recognize immediately from its distinct profile. I’m partial to the Shure Beta 87A supercardioid mic for home recording of audio podcasts, but as you can see from the video below, there are some reasons that you might not want to use it for live video conferences, meetings or events – even if it sounds great even untreated.

There are a range of other options, of course – including differently priced options from both Rode and Sennheiser, most of which offer great quality for what you pay. The nature of audio is that it’s also a highly personal preference, with different people preferring sound that’s either favors the higher end, the low end, or that’s more or less balanced, so it’s going to take a lot of comparison shopping and listening to samples to figure out what works for you.

Bottom line

In the end, sticking to quality brands with established reputations in the film and video industries is a great way to get make the most of your setup. Mics like those I use above benefit even more from physical sound isolation, including measures that are fairly easy to accomplish, like laying down carpets and towels, as well as more advanced practices, like picking up dedicated sound isolating materials including foam pads and mounting them on your walls.

Sound is probably the trickiest part of any videoconferencing or virtual event setup to get right – it’s as much art as it is science, and there are a lot of variables that are hard to control, even with the best equipment, especially in live settings. But going the extra mile can mean the difference between coming across polished and professional, and appearing unprepared, which is bound to make a difference in our increasingly virtual face-to-face world.

Source: Tech Crunch

High quality data is the fuel that powers AI algorithms. Without a continual flow of labeled data, bottlenecks can occur and the algorithm will slowly get worse and add risk to the system.

It’s why labeled data is so critical for companies like Zoox, Cruise and Waymo, which use it to train machine learning models to develop and deploy autonomous vehicles. That need is what led to the creation of Scale AI, a startup that uses software and people to process and label image, lidar and map data for companies building machine learning algorithms. Companies working on autonomous vehicle technology make up a large swath of Scale’s customer base, although its platform is also used by Airbnb, Pinterest and OpenAI, among others.

The COVID-19 pandemic has slowed, or even halted, that flow of data as AV companies suspended testing on public roads — the means of collecting billions of images. Scale is hoping to turn the tap back on, and for free.

The company, in collaboration with lidar manufacturer Hesai, launched this week an open source dataset called PandaSet that can be used for training machine learning models for autonomous driving. The dataset, which is free and licensed for academic and commercial use, includes data collected using Hesai’s forward-facing PandarGT lidar with image-like resolution as well as its mechanical spinning lidar known as Pandar64. The data was collected while driving urban areas in San Francisco and Silicon Valley before officials issued stay-at-home orders in the area, according to the company.

“AI and machine learning are incredible technologies with an incredible potential for impact, but also a huge pain in the ass,” Scale CEO and co-founder Alexandr Wang told TechCrunch in a recent interview. “Machine learning is definitely a garbage in, garbage out kind of framework — you really need high quality data to be able to power these algorithms. It’s why we built Scale and it’s also why we’re using this dataset today to help drive forward the industry with an open source perspective.”

The goal with this lidar dataset was to give free access to a dense and content-rich dataset, which Wang said was achieved by using two kinds of lidars in complex urban environments filled with cars, bikes, traffic lights and pedestrians.

“The Zoox and the Cruises of the world will often talk about how battle-tested their systems are in these dense urban environments,” Wang said. “We wanted to really expose that to the whole community.”

The dataset includes more than 48,000 camera images and 16,000 LiDAR sweeps — more than 100 scenes of 8s each, according to the company. It also includes 28 annotation classes for each scene and 37 semantic segmentation labels for most scenes. Traditional cuboid labeling, those little boxes placed around a bike or car, for instance, can’t adequately identify all of the lidar data. So, Scale uses a point cloud segmentation tool to precisely annotate complex objects like rain.

Open sourcing AV data isn’t entirely new. Last year,  Aptiv and Scale released nuScenes, a large-scale data set from an autonomous vehicle sensor suite. Argo AI, Cruise and Waymo were among a number of AV companies that have also released data to researchers. Argo AI released curated data along with high-definition maps, while Cruise shared a data visualization tool it created called Webviz that takes raw data collected from all the sensors on a robot and turns that binary code into visuals.

Scale’s efforts are a bit different; For instance, Wang said the license to use this dataset doesn’t have any restrictions.

“There’s a big need right now and a continual need for high quality labeled data,” Wang said. “That’s one of the biggest hurdles overcome when building self driving systems. We want to democratize access to this data, especially at a time when a lot of the self driving companies can’t collect it.”

That doesn’t mean Scale is going to suddenly give away all of its data. It is, after all a for-profit enterprise. But it’s already considering collecting and open sourcing fresher data later this year.

Source: Tech Crunch

Minecraft is one of the most popular games on the planet, so it’s natural that Microsoft, after buying creator Mojang some years back, would attempt to apply the genre’s playful, blocky aesthetic to other genres. After modest success with the Story Mode adventure game and Pokémon GO-like Minecraft Earth, they’ve tried their hand at a light action-RPG à la Diablo — and unfortunately come up rather short. For now, that is.

Minecraft Dungeons is a sort of my-first-dungeon-crawler type game, a friendly, streamlined version of the genre Diablo created where players enter a procedurally-created dungeon or region, kill some monsters, get some loot, make it out alive, and do it all over again.

That’s the idea in this game as well, but of course the whole thing uses the block-based look and feel of Minecraft. As you travel through different biomes to free villagers, destroy ancient forges and so on, everything from the levels and monsters to equipment and potions looks like it came straight out of the original game. They nailed the look perfectly.

It’s refreshing, because games like this tend to court a rather grim aesthetic, and when it comes to gameplay they pile on features and mechanics until it feels more like you’re playing a spreadsheet than a game. It’s clear from the start Minecraft Dungeons was intended to provide the fun of fighting, upgrading, and exploring without the overly complex and dark trappings of the genre.

For instance, instead of having a handful of character classes each with their own skill tree, everything your character can do depends on their equipment. Weapons, armor, and accessories all have unique bonuses and abilities. So if you want to be a bow and arrow type fighter, wear the Ranger armor that gives you extra ranged damage and ammo, and use accessories that empower your arrows. Want to be a melee guy? There’s armor and swords for that too.

Customization of your play style, an important part of these games, is achieved by judicious choice of a set of random upgrades on each item. When you gain a level, you get a point can be used to activate, say, a passive ability that deflects enemy projectiles 20 percent of the time. Then it costs two points to upgrade it again, so it deflects 30 percent of the time.

You get those points back when you trash the item and can reapply them to a new one, providing low-risk, low-commitment progress — in time you’ll have lots of points banked to upgrade and experiment with whatever new item you find.

This approach is really a breath of fresh air after the convoluted overlapping systems of the likes of Diablo, Grim Dawn, and Path of Exile. There was just the right amount of “this new sword is tempting but do really I want to recycle my old one?” tension, and although you will collect trash loot, it’s easy to check and dispose of.

I didn’t get a chance to test multiplayer, but the game is definitely designed with co-adventuring in mind. Couch co-op lets you drop in a second player with a controller or connect online with others on the same platform (cross-play is coming soon). A cross-platform casual dungeon crawler is something I’ve been wanting for a long time.

It’s too bad, then, that this is where the game runs out of really positive qualities. I’m keeping in mind that this is a $20 game designed with players new to the genre in mind — not to say kids exactly — so there’s no sense comparing it directly to a major mainstream gaming franchise. But even so, Minecraft Dungeons has some serious issues.

For one thing, it really needs more variety. Part of the fun of these games is traveling from region to region and fighting new types of monsters with different tactics and abilities. That really just isn’t there in this game. The 10 different areas are visually distinct, yes, but they’re linear, similar from one run to another, and don’t differ all that much gameplay-wise. One aspect of Minecraft I’ve always loved, exploration, is nearly absent. Getting up on a hill or down in some little valley or cavern you can see usually isn’t possible — they’re just walls or bottomless pits. Side paths often run quite a distance but I eventually learned to stopped taking them because they were frequently empty and it always took forever to backtrack afterwards.

You’ll run into the same zombies, spiders, and soldiers over and over, and get the same weapons and accessories dropped over and over, often with very similar stats. Although there seems to be a good variety at first, the abilities and weapons don’t seem particularly well balanced, with some obviously and objectively better than others. Some are basically useless: One ability gives you a speedup for a few seconds after you dodge — but the game also slows you down for a few seconds after you roll, so they kind of just cancel each other out. Another returns a third of one percent of your health for every 100 blocks you uncover in the game. What?

This wouldn’t be an issue if the game had better difficulty tuning. I found in my playthrough that there was no challenge whatsoever 99 percent of the time, and then suddenly a situation would arise where I would be nearly instantly killed. These weren’t lesson-teaching deaths like other games — just sudden confluences of bad luck and, it must be said, some poor design.

Ranged attacks from enemies will often come from off-screen, for instance. And not just a stray arrow, but many simultaneously. Enemy projectiles also go through all other enemies, unlike your own, and are very difficult to dodge, especially when there are a dozen coming from different angles. So sometimes after spending the whole level barely taking a hit, you’re reduced to an emergency situation in a fraction of a second, with very little warning, by enemies you haven’t had a chance to react to or perhaps even see. The close-zoom camera shows details well but limits your understanding of what’s happening around you.

These brutal difficulty spikes aren’t always accidental. One enemy kept popping up that repeatedly spawned huge numbers of bear traps under my character’s feet that closed before any but a really expert player could be expected to dodge. Bosses are cheap, swarming players with minions, storms of enormous projectiles, and instant, undodgeable melee attacks.

The issue here isn’t just that it’s hard, but that the game doesn’t give you the tools you need to deal with it. Dodging feels clumsy and enemies block your movement; there is little in the way of active defense like a shield or accessory you activate to repel arrows for 5 seconds; you only have one slowly recharging healing potion and health doesn’t trickle back, so little mistakes add up over time. Not that it matters, since punishment is usually swift and extreme.

What all this amounts to is a game that alternates between monotonous and frustratingly hard, even for a fan of the genre like myself. And considering you’ll run through all the areas in the game in a handful of hours — there are ten areas, each of which takes perhaps 20 minutes to clear — it’s expected that you’ll repeat them over and over to reach the gear level required to beat the final boss. I got all the way to that point and was insta-killed twice in a row.

I repeated a few areas but found them nearly indistinguishable from their earlier iterations. Ultimately I just wasn’t motivated to grind away just so I could unlock another, likely even more unfair, difficulty level.

I wouldn’t complain so much if this wasn’t, ostensibly, a game for beginners. Minecraft Dungeons innovates and simplifies in some really laudable ways, but the moment-to-moment game design is too uneven and the variety on offer isn’t enough even for a $20 game.

But it must be said that Minecraft itself also started out rather barebones and was built up over time into something remarkable and almost infinite. There are two DLC packs in the works for Dungeons, one rather crassly visible from the very start — nothing like being asked to pay more for a game you just bought. The good news is these packs will grow the game to a size that feels more like an adventure and less like a demo. I also expect that patches over the coming weeks and months will considerably tweak the equipment and difficulty — it can be, and needs to be fixed.

A year from now Minecraft Dungeons could very well be a no-brainer purchase, a cross-platform casual hack-and-slash that you can play with your kids or your friends and have a great time without thinking too hard about it (or opening Excel). But right now it’s mostly potential. I’d hold off on picking this one up until it’s been made into the game it’s meant to be.

Source: Tech Crunch