Archiv der Kategorie: Innovation

Apples Rate of Change — The days of enormous iPhone growth may have reached its end

Summary: Apples Rate of Change — the idea that we’ll never see an iPhone sales quarter bigger than this one, or at least not much bigger. The days of enormous iPhone growth may have reached its end.

The reports of the iPhone’s death have been greatly exaggerated.

In the wake of Apple’s recent quarterly financial results report, there’s been a lot of talk about what happens if the company has truly reached the peak of iPhone sales — and what must come next in order for Apple to keep growing.

The iPhone is a once-in-a-decade (if not once-in-a-lifetime) product, and won’t be replaced on Apple’s revenue chart any time soon. And that’s okay, for a whole bunch of reasons.

What goes up… stays up

It’s easy to assume — in part due to language commonly used by growth-obsessed investors — that the iPhone is in free fall. Not so much: iPhone sales set a record last quarter. What’s actually concerning investors is the rate of change — the idea that we’ll never see an iPhone sales quarter bigger than this one, or at least not much bigger. The days of enormous iPhone growth may have reached its end.

If you’re comparing the iPhone’s life cycle to that of Apple’s iPod, there’s reason to be terrified: The iPod sold like gangbusters for a number of years, but its decline was drastic — to the point where it got removed from Apple’s financial reports last year. That’s not going to happen with the iPhone, for a simple reason: the iPod was made largely obsolete by the smartphone. And the smartphone’s not going anywhere, not for a very long time.

That means Apple’s iPhone business is probably going to keep contributing 150 billion dollars a year for the foreseeable future. (In the last four quarters, the iPhone brought in an average of 38.9 billion per quarter. In comparison, the Mac and iPad bring in five or six billion dollars per quarter. That’s a lot of money, sure, but the two products combined pale in comparison to the phone juggernaut.) It’s enough money to make Apple one of the biggest, most profitable companies on the planet.

Could the iPhone eventually fail? The future is promised to no one, but people are going to want an internet-connected device in their pockets until there’s something even better you can stick in your ear or pop on your eyeball or connect directly to your brain.

There’s money in the ecosystem

Apple focused a lot of energy this week on communicating how well it’s doing — and how much it’s growing — in terms of services revenue. That’s the budget line covering iCloud, iTunes, Apple Music, and the App Store.

The users of Apple’s one billion active devices are all spending money on digital goods and services. It’s potentially a huge growth opportunity for the company, and it will be interesting to see what other services Apple might introduce and how much additional revenue might be generated from its existing iPhone installed base — namely, us.

But beyond offering us more content to buy, the Apple ecosystem extends outward. Consider the Apple Watch: It’s essentially an iPhone accessory, since it only works with Apple’s smartphone. It’s another product that can be marketed to existing iPhone users, generating more revenue while also tying them more tightly into the Apple ecosystem. (When an Apple Watch user considers an Android phone, they also have to consider giving up their Apple Watch — making it potentially that much easier to stick with what they know.)

There’s still room for growth

The days of rapid smartphone sales growth may well be over, and while Wall Street may not be thrilled about this, it doesn’t mean the iPhone is in any danger of disappearing. Apple still thinks there’s room for future growth, and the company’s reasons seem reasonable to me. The rapid growth of the middle class in China is creating hundreds of millions of new consumers with money to spend on brands like Apple, and products like the iPhone. Apple’s weak position in India is generally seen as a negative, but it also means there’s a huge upside if the company figures out how to crack that market.

While those of us in the most industrialized nations have benefited from fast 4G LTE cellular networks for a few years now, those networks are still rolling out in India and other emerging markets. People in those countries will buy new phones to take advantage of LTE as it comes online, and that’s a big opportunity for Apple to sell iPhones.

And then there’s switching: Apple continues to suggest that there’s a constant flow of smartphone users from Android to iPhone. It’s hard to quantify those numbers overall, but at least from Apple’s perspective, there’s a growth opportunity simply in picking up Android users who are ready for a change.

Wait for it

Okay, so the iPhone’s pretty good for now. But what about the next big thing? How does Apple ignite future growth, and protect the products it already has?

Fortunately, Apple has many, many billions of dollars in cash from its past few years of profits. And the company is investing that money in researching the next generations of products. I’m sure some of that money is going into exploring what might replace a smartphone, whether it’s a Siri-powered device that plugs into your ear, or an augmented-reality visor, or who knows what else.

In terms of finding growth, we’ve all heard the reports that Apple’s exploring the possibility of building a car. Entering new markets is never easy, but it provides huge opportunity for growth. It’s the same principle as iPhone sales in India: Apple’s current share of the automobile market is zero, which means that the sky’s the limit when it comes to gaining new customers.

The smartphone era

I’m pretty confident that when we look back to the early parts of the 21st century, we will consider this the dawn of the smartphone era. Even from the perspective of 2016, the personal computer seems to rapidly be transforming into a footnote — a technological prelude to the creation of the smartphone. Tiny devices with massive computing power and an always-on connection to a global data network, living in our pockets — they have transformed the way people live around the world, from the richest countries to some of the poorest.

Apple doesn’t need to replicate the iPhone’s success with another product to be successful, which is good, because there may not be a product as successful as the iPhone any time in the near future. (Though I’d be happy to be proven wrong when the direct-brain implants come around in 2030.) People who are searching the horizon for the next big thing as hot as the smartphone are searching in vain.

We live in the smartphone era, and considering the slowing rate of growth in smartphone sales, so does everyone else. The introduction of the iPhone was the moment this era truly began. Apple has benefited massively from that, and will continue to for the foreseeable future.

http://m.imore.com/persistence-iphone

Moving from Brand Loyalty to Experience Loyalty

Huawei at Mobile World Congress 2015 Barcelona

Huawei at Mobile World Congress 2015 Barcelona

Chinese telecoms giant Huawei posted an eye-popping 70% growth in 2015. It is now beating HTC and Sony when it comes to market share in Europe and is third only to Apple and Samsung when it comes to global smartphone sales.

Huawei’s growth is another indication of how Chinese companies are successfully moving away from their traditional strategy of producing cheaper products to attack the low-end of the market.

Huwawei-Ceo-Richard-Yu

The world of smartphones, tablets, smart watches and connected devices of the internet of things is the new battleground to provide digital services. And Huawei has quickly announced itself as a serious player in it. Last year’s success was built on a strategy that rival Western firms have excelled in – marketing, brand building and customer service.

The race to engage

Existing brands such as Blackberry and Sony are already engaged in competitive marketing in both product and customer engagement to grab a piece of the large but finite customer base, so Huawei’s rapid rise to overtake them is very impressive. Huawei increased its global share of the smartphone market from 6.8% in 2014 to 9% in 2015 – a massive 50% gain compared to Apple’s growth of 27%.

It’s a crowded android market. TechStage, CC BY-ND

It suggests Huawei now understands how to play the smartphone market. New organisational and workforce strategies that have a laser focus on a customer-first mindset, tightly driven by the needs of the local market, have been put together. In a crowded price-sensitive android market this matters greatly where choice and brand awareness are critical.

Just competing on product functionality and a brand name is not enough when the nature of “smart” means you have connected consumers and feedback on social media is instant. Huawei understands that the mobile market is now all about the customer service experience and no longer just a telecoms commodity. Getting hold of the consumer and personalising the digital world for them with a good experience and price point that fits their needs and lifestyle choices is also critical.

Huawei is a prime example of a modern commercial mindset emerging from Chinese industries. Its marketing embraces local markets and they aggressively target consumers with sponsorship deals that include a host of big football clubs across Europe, including Arsenal, Paris Saint-Germain and AC Milan. Plus, their product portfolio spans the complete modern telecoms provider from hardware to software.

Huawei is going head to head with Apple. Kārlis Dambrāns, CC BY

Three ways to succeed

Customers cannot be taken for granted in this tough market. History shows that customers lack pure brand loyalty – they are more loyal to the experience, the community and ecosystem of services that best fits their needs. With so much choice out there, it is more about listening and socially engaging with the customer that is key.

At least three key strategies seem to be emerging in the growth of the telecoms players in the new mobile, wearables and connected internet of things services.

  1. A strong focus on “customer first” philosophy from the CEO down through the whole organisation to effectively manager customer experience 24/7.
  2. Using third parties to sell your products and a service that adds value and extends penetration into new markets. Huawei has a partner programme that drives sales across its portfolio. Recent awards in Asia in 2015 follow expansions into Australia in 2011 and similar regional strategies across Latin America, the Middle East and Europe. This federated supply chain model allows companies to extend their workforce and has accounted for more than 55% of Huawei’s growth from these third party sales. It is increasingly essential for scaling up sales regionally.
  3. Embracing international standards to get into thought leadership positions. For example by joining The Open Group, a major software standards consortium, and its Digital Business and Customer Experience (DBCX) workgroup helps Huawei define initiatives on connected customer and product design. Taking part in these kinds of groups enables firms to raise their game and influence in new markets by getting smarter in the way they interact with existing and potential customers and partners. It can then be translated into improving working practices – from managing the supply chain to service delivery – across the board.

What Huawei has done well is realise that consumers are always connected. Companies that exploit this will start to gain more ground in the battle for owning the digital market. It requires more than the specific strategies, but thinking holistically about how to transform to a digital operating model in this new world of connected things.

https://theconversation.com/how-chinas-huawei-is-taking-on-samsung-and-apple-52838

 

teslas self-driving car

Tesla CEO Elon Musk has made a bold prediction: Tesla Motors will have a self-driving car within two years.

“I think we have all the pieces,” Musk told Fortune, “and it’s just about refining those pieces, putting them in place, and making sure they work across a huge number of environments — and then we’re done. It’s a much easier problem than people think it is.”

Although Musk’s comments to Fortune came Monday, The Street pegged a rise in Tesla’s shares to the comments on Tuesday. The ambitious timeframe appeared to be offering support to the stock again today, with shares trading up $1.47, or 0.64 percent, at $231.42 around 7:18 a.m. PST.

Musk’s driverless-car comments may have been overshadowed initially by the achievement of SpaceX on Monday night in landing a rocket during a commercial mission for the first time. Musk is also CEO of SpaceX.

This is the most aggressive timeline Musk has mentioned. While Musk claims the problem is easier than people think it is, he doesn’t think the tech is so accessible that any hacker could create a self-driving car. Musk took the opportunity to call out hacker George Hotz, who claimed via a Bloomberg article last week that he had developed self-driving car technology that could compete with Tesla’s. Musk said he wasn’t buying it.

“But it’s not like George Hotz, a one-guy-and-three-months problem,” Musk said to Fortune. “You know, it’s more like, thousands of people for two years.”

The company went so far as to post a statement last week about Hotz’s achievement.

“We think it is extremely unlikely that a single person or even a small company that lacks extensive engineering validation capability will be able to produce an autonomous driving system that can be deployed to production vehicles,” the company stated. “It may work as a limited demo on a known stretch of road — Tesla had such a system two years ago — but then requires enormous resources to debug over millions of miles of widely differing roads.”

While Tesla is unconcerned about Hotz, the company’s new timeline may have other autonomous car developers hitting the accelerator. Tech companies like Google and Apple, in addition to automakers such as Volvo and General Motors are all competing to be among the first to offer some form of self-driving tech. Many believe the early 2020s would be a realistic timeframe to expect to see the public engaging with self-driving cars.

Just yesterday, it was reported that Google and Ford will enter into a joint venture to build self-driving vehicles with Google’s technology, according to Yahoo Autos, citing sources familiar with the plans. The official announcement is expected to come during the Consumer Electronics Show in January, but there is no manufacturing timeline.

But even if Tesla moves quickly on self-driving cars, are consumers ready for them? The Palo Alto-based carmaker’s recent Firmware 7.1 Autopilot update includes restrictions on self-driving features. The update only allows its Autosteer feature to engage when the Model S is traveling below the posted speed limit. The update came shortly after it was reported that drivers were involved in dangerous activities while the Autopilot features were engaged.

Source: http://www.bizjournals.com/sanjose/news/2015/12/23/elon-musks-bold-new-timeline-for-driverless-cars.html?ana=yahoo

Apple now makes 94% of the profits in the smartphone industry

Source: http://www.businessinsider.de/apple-dominates-profits-by-smartphone-maker-2015-11?r=US&IR=T

Smartphone_Share_2015

Apple now makes 94% of the profits in the smartphone industry, according to recent research by Canacord.

This historical chart compiled by Statista shows how quickly and utterly Apple has dominated the smartphone market. Samsung is now the only other major handset company earning significant profits from smartphones.

Five years ago, the iPhone was still the top profit-maker, but a lot of other companies were in the game. Since then, the platform battle has become a two-player race between Apple’s iOS and Google’s Android, driving third-way competitors like BlackBerry and Microsoft/Nokia down into the loss zone. The fierce competition between Android handset makers, particularly with the rise of inexpensive Chinese Android phones, has also sucked a lot of profit out of the market.

Disrupting automotive through adaptation of technology business model – How to attract MILLENNIALS

n the US 28% of cars are leased. While it is uncommon to lease inexpensive vehicles and family cars, close to half of all luxury cars are. That percentage is only higher in one other car-segment: electric vehicles (EVs): In the first 3 quarters of 2015 75% of new EVs have been leased!

The most common explanation is that EVs are still too expensive to buy. Another popular reason is that customers do not trust the durability of electric powertrains and lithium-ion battery technology. Finally, customers claim that driving range might be an issue and thus prefer leasing over buying (more on my thoughts on driving range anxiety)

All 3 reasons play a major role. All of them have been researched by J.D. Power back in 2010. However, they don’t sufficiently explain the high lease rates among EV customers today. Here are three insights why car leases are 3-4x more common in the EV segment and why car ownership is becoming rare among young customers.

GenY (Millennials) Adapts New Purchasing Habits

Picture

Average Earnings for Young Adults in $2013

Picture

Cars Sold in Millions per Generation

Car leases are already the most popular way of „purchasing“ a luxury and electric vehicle (EV). First, I documented why millennials/younger customers are more likely to lease. Second, I described why technology changes can lead to reduced interest in buying. Finally, I tried to proof that smartphones have given users the ability to experience freedom without owning a car.

These 3 points lead to an assumption: GenY, as the second largest car buying generation, is leading the ownership disruption in the car segment. They buy fewer cars per 1000 citizens, have the highest % of leases and have different expectations for cars (in terms of technologies and features). How can car manufacturers attract GenY and bring driving back?

Lets take a look outside the car industry. How are technology firms attracting young customers? The smartphone market, like the car market, has taken a hit in the last few years. The handset replacement cycle has slowed down significantly. It is the slowest since the introduction of the iPhone in 2007. In 2014, 143 million mobile phones were sold in the United States (-15%). Of them ~90% were smartphones. 2007 users upgraded their phones every ~19 months; today they upgrade every 26+ months.

Picture

Mobile Phone Upgrade Cycle

 

Source: http://www.ev-analyst.com/home/disrupting-automotive-by-adaptation-of-technology-business-model-3-reasons-why-car-ownership-is-dying-12

http://www.ev-analyst.com/home/disrupting-automotive-through-adaptation-of-technology-business-model-how-to-attract-millennials-22

eSIM ab 2016 bei Deutscher Telekom

Quelle: http://www.heise.de/newsticker/meldung/Deutsche-Telekom-eSIM-soll-2016-kommen-2921732.html?wt_mc=nl.ho.2015-11-15

Die klassische SIM-Karte ist am Ende. Laut Telekom wird sie schon ab 2016 von der sogenannte eSIM abgelöst. Diese ist fest in ein Mobilgerät integriert und kann beispielsweise für einen Anbieterwechsel umprogrammiert werden.

Nach 25 Jahren soll die klassische SIM-Karte vom Markt verschwinden. An ihre Stelle tritt schon ab 2016 die sogenannte eSIM, ist sich die Deutsche Telekom sicher. Dabei handelt es sich um eine fest in ein Mobilgerät integrierte, von außen programmierbare SIM-Karte. Das E steht für embedded.

 esim_Deutsche.Telekom


Damit würde der Tausch der SIM-Karte etwa beim Anbieter- oder Gerätewechsel entfallen. Künftig müssten Kunden beispielsweise nur noch den Identifikationscode eines Mobilgeräts einscannen und es so aktivieren. Automatisch werde es dann bereits mit anderen eingebundenen Geräten vernetzt sein, so die Telekom im hauseigenen Blog. Kunden sollen so über einen Vertrag mehr Endgeräte verwalten können als bisher.

Erste Lösungen in Tablets und Wearables

Die Telekom arbeitet nach eigenen Angaben seit Jahren in internationalen Gremien unter dem Dach der GSMA an einem offenen Standard für die eSIM. Er soll die technischen Anforderungen bestimmen und die Regeln für die Profilverwaltung festlegen. „Wir sind überzeugt, dass der neue eSIM Standard ab 2016 in den Markt kommt und sich dann ab 2017 richtig durchsetzt“, heißt es im Blog des Telekommunikationsunternehmens.

Die ersten Lösungen sollen in Tablets und Wearables angeboten werden. Anfangs werde es Hybridlösungen aus eSIM und Plastikkarte geben, so die Telekom weiter. In zehn Jahren werde die klassische SIM dann völlig verschwunden sein.

Apple und Samsung in „fortgeschrittenen“ Verhandlungen

Bereits im Sommer berichtete die Financial Times von Verhandlungen zwischen dem Branchenverband GSMA, Mobilfunkanbietern sowie Apple und Samsung über die Einführung einer umprogrammierbaren SIM für Mobilgeräte. Der Branchenverband zeigte sich seinerzeit optimistisch, eine formelle Vereinbarung mit Apple zu erzielen.

Der US-Konzern führte bereits im vergangenen Jahr die sogenannte Apple SIM ein, die ebenfalls nicht mehr an einen Anbieter geknüpft ist. So sollen Apple-Kunden den Netzanbieter etwa im Urlaub oder auf Geschäftsreisen direkt vom iPad aus wechseln können. Auch in Deutschland ist die Apple SIM mit iPad Air 2 und iPad mini 4 erhältlich. Bisher haben Kunden aber keine große Anbieterauswahl. Zu den Partnern der Apple SIM gehören die Telekom in Form von T Mobile in den USA, EE in Großbritannien sowie GigSky

 

Pass und Führerschein in der Cloud – Zugriff Smartphone

Führerschein kann künftig am Smartphone hergezeigt werden
Führerschein kann künftig am Smartphone hergezeigt werden – Foto: Michael
Nicht im Silicon Valley, sondern in Österreich soll das Problem der sicheren digitalen Identität gelöst werden. Die österreichische Staatsdruckerei hat eine App entwickelt.

Die Frage, wie sich Menschen sicher digital ausweisen können, soll nicht von Facebook, Google und Co gelöst werden. Vielmehr soll österreichisches Know-how dazu beitragen, dass Personalausweise, Reisepässe und Führerscheine künftig nicht mehr in gedruckter Form mitgeführt werden müssen, sondern einfach und sicher auf dem Handy gespeichert werden können.

Ausweis am Smartphone MIA…
Foto: Michael Leitner

Eine entsprechende Lösung inklusive App präsentierte die Österreichische Staatsdruckerei (OeSD) am Donnerstag in Wien. Das System, das auf den Namen MIA („My Identity App“) hört, soll nicht nur Identitätskontrollen in Österreich revolutionieren, sondern kann theoretisch von jedem Staat übernommen werden, der eine digitale Entsprechung zu herkömmlichen Ausweisen umsetzen möchte. Es soll in einem ersten Schritt diversen Staaten weltweit angeboten werden.

System in der Cloud

Um Datenmissbrauch zu verhindern, werden die sensiblen Daten nicht auf dem Mobiltelefon, sondern zentral auf in der Cloud gespeichert, an die das Handy über eine verschlüsselte Internetverbindung andockt. Der Austausch der Daten bei einer Ausweiskontrolle erfolgt erst nach erneuter Bestätigung des Ausweis-Inhabers auf dem eigenen Handy. Das Fremdgerät bekommt dabei über einen generierten Hash bzw. Zifferncode ebenfalls nur Zugriff auf die Cloud – die Daten bleiben dort sicher verwahrt.

Für Behörden – etwa bei Verkehrskontrollen – könnte die Arbeit damit wesentlich erleichtert werden. Zulassungsschein und Führerschein werden inklusive Foto digital übermittelt, aber auch der Unfallort sowie ein Unfallhergang könnten sofort vor Ort elektronisch erfasst werden. Neben der Ausstellung eines Strafzettels in digitaler Form erlaubt das System auch, die Führerscheinberechtigung mit einem Klick zu entziehen oder bei Verdachtsmomenten die Person genauer zu überprüfen.

„Besserer Datenschutz“

Was auf den ersten Blick stark nach gläsernem Mensch klingt, kann in vielen Situationen auch zu einem besseren Datenschutz führen, ist OeSD-Geschäftsführer Lukas Praml im Gespräch mit der futurezone überzeugt. So bekommt ein Türsteher bei der Altersüberprüfung eines Lokalbesuchers lediglich dessen Foto sowie die Bestätigung übermittelt, dass jener über 18 Jahre alt ist. Andere Daten wie Name, genaues Geburtsdatum etc. bleiben bei dieser Ausweiskontrolle geschützt, da sie irrelevant sind.

OeSD Academy Spezial | Produktpräsentation 2015
Lukas Praml präsentierte die neue ID-Lösung – Foto: Österreichische Staatsdruckerei/Martin Hörmandinger

Selbiges kommt auch bei einem privaten Autoverkauf oder beim Ausweisen in einem Hotel zur Anwendung, wo man bisher ohne Zögern seinen Reisepass oder andere persönliche Dokumente aushändigte. Weitere Anwendungsszenarien sind das sichere Einkaufen im Internet oder etwa das Eröffnen eines Bankkontos online inklusive Identitäts-Check. Auch Gesundheitskarten und andere Firmen-ID-Ausweise könnten künftig in MIA hinterlegt werden.Um das System noch sicherer zu gestalten, wird die App zusätzlich mit Fingerprint- oder PIN-Eingabe geschützt. Auch der neue Authentifizierungsstandard FIDO wird bei der Zweifaktor-Umsetzung berücksichtigt. Geht das Handy verloren, können die für das Gerät erteilten Zertifikate gelöscht werden. Die darauf verknüpften Ausweise werden damit für Kriminelle wertlos – laut Praml ein weiterer Vorteil im Vergleich zu gedruckten Ausweispapieren, die nach dem Verlust oftmals in falsche Hände gelangen.

Innenministerium interessiert

Ob Ausweise in Österreich künftig am Handy gespeichert werden, hängt in erster Linie vom Gesetzgeber ab. Innenministerin Johanna Mikl-Leitner begrüßte die Lösung der Staatsdruckerei, die unter anderem für die Herstellung der Reisepässe verantwortlich zeichnet. „Die Chancen der Digitalisierung können nur genutzt werden, wenn wir wissen, mit welchem Gegenüber wir es zu tun haben. Es wäre wünschenswert, dass digitale Identitäten so sicher wie der Reisepass sind und solche Systeme auch in anderen Staaten zum Einsatz kommen“, sagte Mikl-Leitner.

MIA OeSD…
MIA auf der Apple Watch – Foto: /Martin Stepanek

Laut Praml ist das System technisch in einem halben Jahr umsetzbar. Bis die rechtlichen Rahmenbedingungen sowie Fragen zu Datenschutz und Kosten geklärt sind, dürfte es aber noch länger dauern. Die App ist für Android und iOS praktisch fertig entwickelt, auch Smartwatches auf Android-Wear-Basis sowie die Apple Watch werden mit adaptierten App-Versionen unterstützt. Wer bei der Entwicklung von MIA auf dem laufenden bleiben will, kann sich auf der Webseite der Staatsdruckerei per E-Mail eintragen.

Quelle: futurezone.at/digital-life/pass-und-fuehrerschein-kuenftig-auf-dem-handy/163.623.515

Google Creates Terminator Like Email Response System

Source: http://googleresearch.blogspot.co.at/2015/11/computer-respond-to-this-email.html

possible-response

Google Creates Terminator 2 Like E-Mail Response System and details the functionality:

Machine Intelligence for You

What I love about working at Google is the opportunity to harness cutting-edge machine intelligence for users’ benefit. Two recent Research Blog posts talked about how we’ve used machine learning in the form of deep neural networks to improve voice search and YouTube thumbnails. Today we can share something even wilder — Smart Reply, a deep neural network that writes email.

I get a lot of email, and I often peek at it on the go with my phone. But replying to email on mobile is a real pain, even for short replies. What if there were a system that could automatically determine if an email was answerable with a short reply, and compose a few suitable responses that I could edit or send with just a tap?

Some months ago, Bálint Miklós from the Gmail team asked me if such a thing might be possible. I said it sounded too much like passing the Turing Test to get our hopes up… but having collaborated before on machine learning improvements to spam detection and email categorization, we thought we’d give it a try.

There’s a long history of research on both understanding and generating natural language for applications like machine translation. Last year, Google researchers Oriol Vinyals, Ilya Sutskever, and Quoc Le proposed fusing these two tasks in what they called sequence-to-sequence learning. This end-to-end approach has many possible applications, but one of the most unexpected that we’ve experimented with is conversational synthesis. Early results showed that we could use sequence-to-sequence learning to power a chatbot that was remarkably fun to play with, despite having included no explicit knowledge of language in the program.

Obviously, there’s a huge gap between a cute research chatbot and a system that I want helping me draft email. It was still an open question if we could build something that was actually useful to our users. But one engineer on our team, Anjuli Kannan, was willing to take on the challenge. Working closely with both Machine Intelligence researchers and Gmail engineers, she elaborated and experimented with the sequence-to-sequence research ideas. The result is the industrial strength neural network that runs at the core of the Smart Reply feature we’re launching this week.

How it works

A naive attempt to build a response generation system might depend on hand-crafted rules for common reply scenarios. But in practice, any engineer’s ability to invent “rules” would be quickly outstripped by the tremendous diversity with which real people communicate. A machine-learned system, by contrast, implicitly captures diverse situations, writing styles, and tones. These systems generalize better, and handle completely new inputs more gracefully than brittle, rule-based systems ever could.

Diagram by Chris Olah

Like other sequence-to-sequence models, the Smart Reply System is built on a pair of recurrent neural networks, one used to encode the incoming email and one to predict possible responses. The encoding network consumes the words of the incoming email one at a time, and produces a vector (a list of numbers). This vector, which Geoff Hinton calls a “thought vector,” captures the gist of what is being said without getting hung up on diction — for example, the vector for „Are you free tomorrow?“ should be similar to the vector for „Does tomorrow work for you?“ The second network starts from this thought vector and synthesizes a grammatically correct reply one word at a time, like it’s typing it out. Amazingly, the detailed operation of each network is entirely learned, just by training the model to predict likely responses.

One challenge of working with emails is that the inputs and outputs of the model can be hundreds of words long. This is where the particular choice of recurrent neural network type really matters. We used a variant of a „long short-term-memory“ network (or LSTM for short), which is particularly good at preserving long-term dependencies, and can home in on the part of the incoming email that is most useful in predicting a response, without being distracted by less relevant sentences before and after.

Of course, there’s another very important factor in working with email, which is privacy. In developing Smart Reply we adhered to the same rigorous user privacy standards we’ve always held — in other words, no humans reading your email. This means researchers have to get machine learning to work on a data set that they themselves cannot read, which is a little like trying to solve a puzzle while blindfolded — but a challenge makes it more interesting!

Getting it right

Our first prototype of the system had a few unexpected quirks. We wanted to generate a few candidate replies, but when we asked our neural network for the three most likely responses, it’d cough up triplets like “How about tomorrow?” “Wanna get together tomorrow?” “I suggest we meet tomorrow.” That’s not really much of a choice for users. The solution was provided by Sujith Ravi, whose team developed a great machine learning system for mapping natural language responses to semantic intents. This was instrumental in several phases of the project, and was critical to solving the „response diversity problem“: by knowing how semantically similar two responses are, we can suggest responses that are different not only in wording, but in their underlying meaning.

Another bizarre feature of our early prototype was its propensity to respond with “I love you” to seemingly anything. As adorable as this sounds, it wasn’t really what we were hoping for. Some analysis revealed that the system was doing exactly what we’d trained it to do, generate likely responses — and it turns out that responses like “Thanks“, „Sounds good“, and “I love you” are super common — so the system would lean on them as a safe bet if it was unsure. Normalizing the likelihood of a candidate reply by some measure of that response’s prior probability forced the model to predict responses that were not just highly likely, but also had high affinity to the original message. This made for a less lovey, but far more useful, email assistant.

Give it a try

We’re actually pretty amazed at how well this works. We’ll be rolling this feature out on Inbox for Android and iOSlater this week, and we hope you’ll try it for yourself! Tap on a Smart Reply suggestion to start editing it. If it’s perfect as is, just tap send. Two-tap email on the go — just like Bálint envisioned.

Terminator Response

This blog post may or may not have actually been written by a neural network.