14
Music
Trailers
DailyVideos
India
Pakistan
Afghanistan
Bangladesh
Srilanka
Nepal
Thailand
StockMarket
Business
Technology
Startup
Trending Videos
Coupons
Football
Search
Download App in Playstore
Download App
Best Collections
Technology
Barcelona-based online travel agency Exoticca— which sells &affordable luxury& holidays to popular destinations such as India, Kenya, Brazil, Thailand and South Africa — has closed a€3.5 million (~$4.1 million) Series A to expand into more markets.
The lead investor is early-stage Madrid-based VCK Fund, with existing investors Sabadell Venture Capital and Grupo Palau also participating, along with new investors Nero Ventures, Palladium Corporate Venture and Smartech Capital.
Exoticca was founded in 2013 and currently operates in three European markets: Spain, France and the U.K. The new funding will be put toward expanding that tally — with the German market next in line, and a launch into the U.S. and Canada also on the horizon.Funds will also be funneled into further developing the platform.
&The company spent the first couple of years developing the technological platform and sales have grown very rapidly since then(€4.4 million in 2016, €10.5 million in 2017 and a budget of more than €20 million for 2018),& saysCEO Pere Valles, a recent recruit to the business and previously CEO of Scytl.
Valles argues that Exoticcaprogress to date proves both the profitability of its business model — noting that Spain was &the first market which Exoticca launched is already profitable& — and its replicability. &Last year we launched the U.K. and France and the U.K. is already bigger than Spain,& he says, adding: &In July, we are launching in Germany and we have plans to open in the U.S. and Canada in 2019.&
Valles says the market Exoticca is operating inis one of the few travel market segments that has not yet been digitized — with people still purchasing these types of trips in traditional &offline& travel agencies, owing to relative complexity, with the holidays typically having multiple legs and components, perhaps including international and domestic flights, land transportation, hotels in different locations, tour guides and so on.
Exoticcaplatform allows users to buy such trips online in a single visit, thanks to a proprietary booking engine that integrates with all the various providers — enabling real-time pricing for each component (so no need to phone up for an actual price before being able to book, for example).
&There is nobody else who provides real-time pricing for these types of trips through an online platform,& argues Valles. &Our competition uses internet only to generate leads and then close the sale either on the phone or in a store while we allow our customers to do the entire purchase process online in a single visit.&
There are some differences versus traditional bricks-and-mortar travel agents, though. Exoticca customers can&t spec out a very bespoke holiday in discussion with an agent, for example.
Rather it offers an inventory of around 50 trip packages in its permanent portfolio, covering what are described as &the most popular destinations& for its target travel category. (Though Valles points out it does offer a degree of light personalization — such as being able to pick a hotel category and optional excursions, for example.)
If you&re content to choose from the selection, Exoticca claims the trips are30 percent cheaper on average versus &traditional providers& — as a consequence of the disintermediation process (i.e. it acting as both wholesaler and retailer).
&Each one of these trip packages is our ‘own& product in the sense that we are the ones who ‘build& it by engaging directly with the provider of each component,& says Valles, adding: &We also give our own personal touch to the tours in each one of these destinations.&
Therea pretty striking branding style on show too — which features 1950s-esque graphics illustrating elements of the holiday packages and service…
Presumably the hope is the retro styling will resonate with the older adults who are the demographic most likely to be in the market for long-haul, luxury trips.
&We have customers in all age groups but those between 45 and 65 tend to be ‘overrepresented,& & agrees Valles.
He says the company is generally targeting a similar customer profile to that of GV-backed members-only travel club Secret Escapes.
Though they are not like-for-like competitors, with Exoticcaproduct certainly having more of a focus on, well, exotic holidays — versus Secret Escapes offering hotel getaways to almost anywhere (so long as the hotel is up to snuff).
Other European online travel agency startups include the likes ofDreamlines, which is focused exclusively on cruise holidays to address a distinct market segment; and Evaneos, a marketplace for tailored travel experiences that connects travelers directly with a community oflocaltravel agents — so is doing the lead generation Exoticcaapproach avoids.
Valles says the packages it sells are with &high-quality providers& (4- and 5-star hotels) but offered at &discounted prices& intended to appeal to a mass market of middle- and upper-class travelers.
The overall aim is to &democratize& this segment of the travel market. (&Agreat experience at a reasonable cost& is the pitch.) Though if they really succeed in widening the funnel they may end up undermining their luxury promise. But clearly thatnot something they have to worry about yet.
Funding wise, Valles says Exoticca previously raised€1 million in two seed rounds, one with F-F and another with Sabadell Venture Capital. The business is not breaking out any user or usage metrics at this stage, five years in.
- Details
- Category: Technology
Read more: Online travel agency Exoticca bags $4.1M for market expansion
Write comment (96 Comments)Egnyte has always had the goal of protecting data and files wherever they live, whether on-premises or in the cloud. Today, the company announced a new feature to help customers comply with GDPR privacy regulations that went into effect in Europe last week in a straight-forward fashion.
You can start by simply telling Egnyte that you want to turn on &Identify sensitive content.& You then select which sets of rules you want to check for compliance including GDPR. Once you do this, the system goes and scans all of your repositories to find content deemed sensitive under GDPR rules (or whichever other rules you have selected).
Photo: Egnyte
It then gives you a list of files and marks them with a risk factor from 1-9 with one being the lowest level of risk and 9 being the highest. You can configure the program to expose whichever files you wish based on your own level of compliance tolerance. So for instance, you could ask to see any files with a risk level of seven or higher.
&In essence, ita data security and governance solution for unstructured data, and we are approaching that at the repository levels. The goal is to provide visibility, control and protection of that information in any in any unstructured repository,& Jeff Sizemore, VP of governance for Egnyte Protect told TechCrunch.
Photo: Egnyte
Sizemore says that Egnyte weighs the sensitivity of the data against the danger it could be exposed and leave a customer in violation of GDPR rules. &We look at things like public links into groups, which is basically just governance of the data, making sure nothing is wide open from a file share perspective. We also look at how the information is being shared,& Sizemore said. A social security number being shared internally is a lot less risky than a thousand social security numbers being shared in a public link.
The service covers 28 nations and 24 languages and itpre-configured to understand what data is considered sensitive by country and language. &We already have all the mapping and all the languages sitting underneath these policies. We are literally going into the data and actually scanning through and looking for GDPR-relevant data thatin the scope of Article 40.&
The new serviceis generally available on Tuesday morning. The company will be makign an announcement at the InfoSecurity Conference in London. It has had the service in Beta prior to this.
- Details
- Category: Technology
Read more: Egnyte releases one-step GDPR compliance solution
Write comment (90 Comments)Copyright is a mess, but more of a mess in some ways than others, and one of the biggest messes right now is licensing music for digital broadcast. The Music Modernization Act aims to smooth over some of the biggest bumps, but a companion piece has aroused the ire of a collection of internet advocacy groups, which have voiced their concerns via a letter penned by copyright scholar Lawrence Lessig.
The CLASSICS Act is the one in question, though no one is disputing the cleverness of its acronym: it stands for Compensating Legacy Artists for their Songs, Service, and Important Contributions to Society. And indeed the goal of the act is to harmonize copyright for musical works created before 1972, a sort of turning point after which copyright law changed considerably.
The issue at hand, although due to the complexity of copyright any summary will necessarily be somewhat inadequate, is that pre-1972 works are essentially ineligible for federal copyright status, and are instead registered by a hodgepodge of state and common law. This was a pain to start with, to be sure, but as digital transmission of music has increased and federal copyright become more important, it has really started to chafe and the question of royalties has become an increasingly complicated one. (Plagiarism Today has a good summary if you want more details.)
What CLASSICS is intended to do is endow those pre-1972 works with a semblance of federal copyright, helping eliminate this arbitrary line on how music is licensed. Sounds good, right And it mostly is. But therea catch — or perhaps two catches.
&The objective of the whole package is to make much more efficient the process of licensing music in the digital age, and I think thatgreat,& Lessig told me in an interview. &But I think, and this is an important line to draw, what we&re criticizing is that the public doesn&t benefit.&
He says the issues came up in discussions with other IP experts and in particular with Brewster Kahle, founder of the Internet Archive, who Lessig said had faced serious legal obstacles relating to music archives in particular.
&Americacopyright system needs to be modernized, but not like this,& the letter at Equal Citizens reads, &and certainly not in a fashion that results in excessively long periods of protection, creates massive inconsistencies in the copyrighting of sound recordings, and gives inequitable access to those sound recordings.&
Co-signatories include Demand Progress, the EFF, Public Knowledge, and the Internet Archive.
A century and a half of protection
First there is the issue that under CLASSICS, the pre-1972 works are given what amounts to (though is legally not quite the same as) a copyright extension covering them until 2067. Why 2067 That1972 plus 95 years, the duration a work would be copyrighted if registered today.
But itnot just works from 1972 that will be covered up to 2067. Works going back another 50 years will also be protected until 2067. In other words, a 1927 recording would be getting what amounts to 140 years of copyright protection.
In case you don&t see the harm here, consider that such a recording would normally have entered the public domain a handful of years from now, but under CLASSICS would be granted another 42 years, during which royalties would have to be paid and rights managed.
Now, the recurring extension of copyrights over years tends to benefit rights holders more than consumers. At the same time, ithard to imagine something like Mickey the Mouse being public domain, but really, there has to be some limit. And for a work to be kept out of the public domain for nearly a century and a half seems fundamentally excessive.
Brewster Kahle told me in an email that CLASSICS &would not release even Edison wax cylinders into the public domain until 2067 — nuts, if you ask me.&
Dozens of scholars argued this point in a letter issued in May:
The Act is deeply flawed…Because it grants new federal protections only to works that were created long ago (ranging in age from 46 to 95 years), it does nothing to incentivize the creation of new works. Rather, it simply provides new rewards to existing copyright owners.
Critics point out that the works aren&t actually being given copyright extensions — they&d be governed under a different law. But that has the air of semantics: CLASSICS itself assigns the same &remedies& as copyright infringement as punishment for the same crime. Itnot copyright, but it sure looks, walks and would be prosecuted like copyright. Ita sort of white-label copyright, since extending and reducing actual copyright terms is a bit of a difficult issue.
Retroactive without registration
But the 2067 thing is just more of the same copyright extension stuff we&ve seen, Lessig says.
&A lot of people are concerned about the length, and I am too,& he told me. &But the 2067 thing doesn&t matter as much as the idea that you have to do something to earn this right. If you&re going to give the extension or benefits, at least do it in a way that clarifies the rights to the public.&
The main issue is that CLASSICS would apply universally to every work created before 1972, regardless of age, current status, whether the creators are alive and still benefiting and so on. A hit song played daily on oldies stations is given the same rights as an obscure jazz recording no one has ever attempted to monetize or reissue.
The latter works are the type of thing that should be entering the public domain, Lessig and his colleagues argue. Under CLASSICS they are copyright protected by default, and their public domain status delayed by decades when they should be available for free like other old and abandoned media.
&I&d like to see a clear line between works people use and works people have no use for,& Lessig said.
Let the likes of The Temptations benefit from their hit songs for another century, if thatwhat the government and industry decide is a reasonable term — all they should have to do is register those works.
By having artists and rights owners register, it solves the problem for everyone. Anyone who wants to have their pre-1972 works brought into the new scheme can easily achieve that, but orphan works will enter the public domain as they ought to.
This has the side effect of centralizing and formalizing the copyright status of pre-1972 works; if you&re a documentary producer or DJ looking for some music to use in your own project, you&ll be able to easily check if some &60s record is rights-managed, in which case you can license it, or if itan orphan, in which case you can safely use it like any other public domain work.
Registration has been a sticky issue in IP law, as it isn&t necessary for the creation of a copyright — but as this isn&t creation of copyright, in fact not technically copyright at all in some ways, itan ideal time to introduce the concept to the benefit of the public and with very little if any inconvenience to rights holders.
Music industry coalition musicFIRSTexecutive director, Chris Israel, disagrees: calling registration &a terrible idea,& he suggested it was onerous for artists who are already dealing with the pre/post-1972 complexities to have to &jump through new legal hoops.& He was not concerned that many works scheduled to enter the public domain would not do so.
&I&m concerned about getting protection for the thousands of valuable pre 72 performances being exploited commercially, not exempting them. Letstart with that,& he wrote in an email.
Arguments for against are necessarily speculative, but all the same thereno reason to think that the registration process would be complex, time-consuming, or labor-intensive. Itreally just a basic filter to separate used from unused works and could even be extracted from some other part of the overall transition process.
&We&ve never had a point to leverage this point into the system,& Lessig said. Therestill time to add it and it would cost little to do so, he added.
Collective, automatic rights managed at a high level are the domain of large companies that can navigate the agreements and pay the huge fines associated with operating outside their terms while the law catches up with the rest of the world. CLASSICS, Lessig suggested, would benefit these companies.
&It concentrates their power as a sort of clearing house for intellectual property rights,& he said. &Right now we have no way to tell what is protected, which has what rights and why. The people that can do it are companies like Spotify and Google, that can afford the liability. But that doesn&t make the culture that nobody cares about available to the public.&
(NB: Lessig, among others, has been accused of being a paid academic puppet for Google; he laughed at the suggestion, saying &I&d love to see one of those checks.& Many of said scholars have similarly rejected the notion, though certainly private funding in science must be closely monitored.)
Kahle and others have recommended as an alternative to CLASSICS or even a modified version thereof, a rival bill called ACCESS, which takes a simpler approach to federalizing the copyrights of pre-1972 works. Perhaps too simple, critics argue: it would put a lot of work directly in the lap of the Copyright Office, which, like most federal organs, is overworked and inefficient enough as is.
The Music Modernization Act has received remarkable bipartisan support; it passed the House 415-0, which is nothing short of incredible in this day and age. So it seems to be on track to becoming a law. But that doesn&t mean itperfect, and just as we have ended up kicking ourselves for sloppy wording and a lack of foresight in parts of the Digital Millennium Copyright Act, it might be better to take a moment to get it right the first time.
- Details
- Category: Technology
Read more: Advocacy groups knock ‘unjust’ copyright-extending CLASSICS Act
Write comment (95 Comments)
Listen up, early-stage startup founders! Do you know the line that Marlon Brando made famous in the classic film, &On the Waterfront& Well, if you don&t act within the next 48 hours and apply to compete in Startup Battlefield, the phrase &I coulda been a contender& might haunt you for a very long time.
TechCrunch Disrupt San Francisco 2018— which takes place September 5-7 — is going big this year. We&re taking over Moscone Center West, tripling our floor space and playing host to more than 10,000 attendees, 1,200 exhibitors and more than 400 accredited media outlets. With such a <cough> disruptive Disrupt, no &ordinary& Startup Battlefield would do. So, we doubled the money to a very chill $100,000 equity-free cash prize.
You have absolutely nothing to lose by applying and a lot to gain if your company makes the cut. TechCrunch charges no fees and takes no equity from startups; applying to and competing in Startup Battlefield is 100 percent free. The selection process is highly competitive — savvy TechCrunch editors review every application and choose anywhere from 15-30 startups. The acceptance rate ranges between three and six percent.
But herethe thing: Just making the first cut comes with a hefty load of benefits in the form of investor interest, media coverage and free exhibition space in Startup Alley — a birthing ground for magical wheeling and dealing — for all three days of the show.
Consider, for example, the case of Aircall, a cloud-based call center solution that competed in Startup Battlefield SF back in 2015. It didn&t move beyond round one but, trust us, they aren&t crying about it. Probably because Aircall recently scored $29 million inanother round of funding. That brings its total funding to $40.5 million since its Battlefield debut.
Hereanother bennie at which you ought not sneeze. Every competing team joins the ranks of the Startup Battlefield alumni community.Imagine the connections you can make among the more than 800 companies in this elite cohort. Companies that have, by the way, collectively raised more than $8 billion in funding and produced more than 100 exits.
If you suffer from performance anxiety, we&ve got you covered on that score. Every Startup Battlefield team receives free expert pitch coaching from our experienced, seen-it-all TechCrunch editorial team. We&ve been doing this since 2007, folks. You WILL be prepared to step into the ring to give it your best shot.
Disrupt San Francisco 2018takes place September 5-7 at Moscone Center West. Your chance to apply and be a contender disappears in just two short days. Steer clear of Palooka-ville. Apply to Startup Battlefield right now.
- Details
- Category: Technology
Read more: Two days left to apply for Startup Battlefield at Disrupt SF 2018
Write comment (100 Comments)Apple annual developer conference, WWDC, started this afternoon down in San Jose — kicking off with a keynote as it does every year where it announced a bunch of updates to all of its major operating systems. We&ve wrapped up all the big announcements from its keynote below, and there will be plenty more information to come in over the next few days.
Apple introduces iOS 12
What Apple announced: Its next generation iPhone operating system. iOS 12 comes with a variety of new features, many of which we&ll dig into below, but it includes things like group FaceTime, new Animoji, some quality of live improvements to Siri, improvements in performance (especially for older devices) and updates to its push into augmented reality.
Why it matters:WWDC has always been the place where it&ll show off a slew of new features, which while very consumer-y in feel is typically pointed at developers to show all the tools they&ll soon get their hands on and all the reasons Apple will keep users on iPhones. The company is introducing its own file format for augmented reality, clearly signaling to developers that they should be flocking to iOS 12 if they want to go where the users are in the future.
Apple delivers big updates to its augmented reality platform
What Apple announced: Apple said it is rolling out ARKit 2.0, the next generation of its development software for augmented reality experiences on the iPhone. The new version showed improved face tracking, more realistic rendering, 3D object detection, persistent experiences and shared experiences (as well as a neat party trick when it comes to measurement). Therealso a multiplayer component the company will roll out with its new development tools.
Why it matters: Tim Cook has alluded to the importance of augmented reality for the iPhone multiple times over the years. While it was a sort of promise of the future for a while, the release of Pokémon Go and the phenomenon that followed demonstrated the potential of augmented reality to capture the hearts and minds of potential users. That was just one use case of what could be a massive potential app ecosystem, and Apple wants to get ahead of that with robust development kits that will keep users on iPhones with better augmented reality experiences.
With iOS 12, Apple focuses on performance
What Apple announced: Apple loves to talk about how its new OS generations, including iOS 11, have the highest adoption rate among smartphones — as well as show off a graphic that shows how bad Google is at that with Android. But it spent a lot of time today talking about how efficient its next OS, iOS 12, will be on older phones like the iPhone 6.
Why it matters: Apple can&t leave its older users in the dust. While itbetter that they upgrade to newer phones, some users are sitting on older devices like an iPhone 6S, and rolling out new operating systems with more robust and complex features might put a strain on those older devices. Apple got in trouble for not notifying users that it was slowing phones with older batteries down, and if it wants to keep people in the ecosystem — and eventually upgrade — it still has to keep those users with older phones excited about the Apple ecosystem amid a ton of competition.
AppleMemoji brings an animated you to your iPhone
What Apple announced: If you&ve ever used Bitmoji, you know exactly what to expect here. Apple is giving users a way to create a customized avatar for themselves that will behave exactly like an Animoji, its animated emoji that moves around as you move your head. Plus you can stick out your tongue and your Animoji will do the same, for some reason.
Why it matters: Snap spent more than $60 million on the makers of Bitmoji. Clearly this is a feature that users want, and itstarting to show up in a lot of different ways as an attempt to differentiate a communications platform. Apple obviously needs iMessage to succeed because it continues to create iOS lock-in, and having these kinds of customized avatars can make the experience more robust.
Apple is adding group FaceTime video calls to iOS 12
What Apple announced: Apple is adding group FaceTime video calls to iOS 12, where you can chat with up to 32 people.
Why it matters:Apple is adding group FaceTime video calls to iOS 12, where you can chat with up to 32 people! This is an interesting and obvious move for Apple to port over the capabilities of Houseparty, an app that tapped a weird zeitgeist around multi-user video streaming. Managing that many streams is difficult and compute intensive, so it makes sense that Apple could absorb the shock of the challenges and bake that right into iOS.
Apple introduces watchOS 5
What Apple announced:Apple showed a bunch of new features that will show up in the next generation of the operating system for the Apple Watch. That includes new features like new workout types like yoga and hiking, challenges for friends, and automatic workout detection. Therealso Siri shortcuts and the walkie talkie, which we&ll get to below.
Why it matters: The most significant of these announcements focused in the health realm, which is where Apple is increasingly positioning itself with the Apple Watch. Originally seen as a sort of do-it-all accessory, it turns out the whole wearable category doesn&t really click for that, but it makes a lot of sense as a fitness tracker and a way to monitor health. Itstill an important part of Appleecosystem, and creating better experiences around workouts can help the company position the Watch as the best fitness tracker on the market.
Apple unveils new screen time controls for childrenand a new set of ‘digital wellness& features for better managing screen time
What Apple announced: Apple is adding some options in iOS to track usage of certain apps, as well as add time limits to flag users when they&re approaching the self-imposed boundaries within iOS. The updates also include more robust do-not-disturb modes. All this also extends to parental controls for children.
Why it matters: Apple is announcing this just after Google unveiled a slew of updates to its new Android operating system that were also focused on digital wellness. Itbecome an increasing focus for the creators of the operating systems to try to discourage users from just tapping around and wasting time on some apps — and hopefully promote better behavior, which would make the whole experience nicer (and, of course, get them to buy new phones). The parental controls part is also significant given that investors questioned Apple when it came to screen time for children.
Apple gives users control of Siri with new Shortcuts tool
What Apple announced: Apple is giving its users the ability to create custom commands with Siri. The whole process involves chaining together a bunch of activities and queries within Siri that users can piece together to respond to a single voice command like &I&m headed to the gym.&
Why it matters: Siri is widely considered to be a much weaker service compared to Alexa or Google Assistant, and it certainly seems like something Apple is not ignoring. The company argues it works to protect privacy, but that comes at a cost, and Apple has to find a way to ensure that its voice assistant is competitive with other products out there.
The next version of macOS is macOS Mojave
What Apple announced: The companynext version of its Mac operating system. Mojave, its latest update, brings in a ton of incremental updates for the service that include a &dark mode& that dims most of the elements on the screen. There are also a bunch of new tools to help users stay a little more organized, such as a new way of viewing files in Finder and stacking documents intelligently.
Why it matters: While the Mac operates a smaller niche inside Applelarger business — which, to be clear, is driven by phones — it has to keep those Mac users happy. They can be among Applemost avid fans, and the Mac serves as another piece of Appleoverall ecosystem that sits alongside the phone and tablet. If Apple wants to pitch additional devices like the Watch or the HomePod, it has to convince users to stay within the Apple ecosystem. That means ensuring its laptop is up to date with new features every year.
Apple Watch gets Walkie-Talkie mode
What Apple announced: You can talk into your watch like a walkie talkie.
Why it matters: You can talk into your watch like a walkie talkie. Some people at TechCrunch care about that a lot for some reason. Itanother thing Apple Watch users have to play with that might get them to buy more watches. Or not.
Apple aims to simplify the Mac App Store with a redesignWhat apple announced: AppleMac App Store, it&sother App Store for its line of laptops and computers, is getting a complete overhaul. Everything is divvied up into tabs and more intelligent grouping, and Apple is making it easier for developers to prompt users to rate their apps.
Why it matters: Apple launched the Mac App Store years ago, but it hasn&t seen any major updates since Apple began making some significant changes to the Apple App Store. The company has taken more of an editorial bent for the App Store, looking to surface up the best apps in an era where the App Store is getting increasingly crowded. So it makes sense that the company would port over those learnings to the Mac App Store.
App Store hits 20M registered developers and $100B in revenues, 500M visitors per week
What Apple announced:The numbers above.
Why it matters: Apple loves to tout these numbers every year, but italso an important barometer for the success of the Apple App ecosystem. Itkind of like looking at a stock chart — you might hear a company is doing well when you&re deciding whether or not to invest in something, but itgood to have that nice round public-facing number.
Apple TV gets Dolby Atmos and streamlined sign-ons for channels and servicesWhat Apple announced: The company is making it easier to sign on and also introducing Dolby Atmos audio, two quality-of-life improvements for Apple TV owners. The latter of which is a nice way to keep the experience clean, but the former makes the Apple TV the only streaming device to be both Dolby Atmos and Vision certified.
Why it matters: Apple insists that the TV is not a hobby, so it keeps bringing these updates to its living room device. While it really hasn&t gottenthere yet, owning the living room is a tantalizing piece of the home puzzle that would help Apple not only sell more devices, but keep people locked into its ecosystem. That&ll be especially true as more and more internet-connected devices end up in the home, all of which needing some kind of hub — which could be the Apple TV.
Apple says CarPlay will now support third party navigation and mapping apps
What Apple announced: CarPlay, the companycar-focused operating system, will support third-party apps like Google Maps and Waze.
Why itimportant: This is pretty self-explanatory. Ita nice quality of life improvement that might make users a little more interested in using CarPlay. Apple doesn&t have a true car play yet, so to speak, but this is one way to start getting users accustomed to iOS in the car.
- Details
- Category: Technology
Read more: 14 big announcements from Apple’s annual developer conference WWDC 2018
Write comment (95 Comments)Apple announced a new feature for developers today called Create ML. Because machine learning is a commonly used tool in the developer kit these days, it makes sense that Apple would want to improve the process. But what it has here, essentially local training, doesn&t seem particularly useful.
The most important step in the creation of a machine learning model, like one that detects faces or turns speech into text, is the &training.& Thatwhen the computer is chugging through reams of data like photos or audio and establishing correlations between the input (a voice) and the desired output (distinct words).
This part of the process is extremely CPU-intensive, though. It generally requires orders of magnitude more computing power (and often storage) than you have sitting on your desk. Think of it like the difference between rendering a 3D game like Overwatch and rendering a Pixar film. You could do it on your laptop, but it would take hours or days for your measly four-core Intel processor and onboard GPU to handle.
Thatwhy training is usually done &in the cloud,& which is to say, on other peoplecomputers set up specifically for the task, equipped with banks of GPUs and special AI-inclined hardware.
Create ML is all about doing it on your own PC, though: as briefly shown onstage, you drag your data onto the interface, tweak some stuff and you can have a model ready to go in as little as 20 minutes if you&re on a maxed-out iMac Pro. It also compresses the model so you can more easily include it in apps (a feature already included in Apple ML tools, if I remember correctly). This is mainly possible because itapplying Appleown vision and language models, not building new ones from scratch.
The quality of a model depends in great part on the nature, arrangement and precision of the &layers& of the training network, and how long itbeen given to cook. Given an hour of real time, a model trained on a MacBook Pro will have, letjust make up a number, 10 teraflop-hours of training done. If you send that data to the cloud, you could choose to either have those 10 teraflop-hours split between 10 computers and have the same results in six minutes, or after an hour it could have 100 teraflop-hours of training, almost certainly resulting in a better model.
That kind of flexibility is one of the core conveniences of computing as a service, and why so much of the world runs on cloud platforms like AWS and Azure, and soon dedicated AI processing services like Lobe.
My colleagues suggested that people who are dealing with sensitive data in their models, for example medical history or x-rays, wouldn&t want to put that data in the cloud. But I don&t think that single developers with little or no access to cloud training services are the kind that are likely, or even allowed, to have access to privileged data like that. If you have a hard drive loaded with the PET scans of 500,000 people, that seems like a catastrophic failure waiting to happen. So access control is the name of the game, and private data is stored centrally.
Research organizations, hospitals and universities have partnerships with cloud services and perhaps even their own dedicated computing clusters for things like this. After all, they also need to collaborate, be audited and so on. Their requirements are also almost certainly different and more demanding than Appleoff the shelf stuff.
I guess I sound like I&m ragging for no reason on a tool that some will find useful. But the way Apple framed it made it sound like anyone can just switch over from a major training service to their own laptop easily and get the same results. Thatjust not true. Even for prototyping and rapid turnaround work it doesn&t seem likely that a locally trained model will often be an option. Perhaps as the platform diversifies developers will find ways to make it useful, but for now it feels like a feature without a purpose.
- Details
- Category: Technology
Read more: Apple’s Create ML is a nice feature with an unclear purpose
Write comment (92 Comments)Page 5171 of 5614