Blog

ProjectBlog: Mobile — Edge of Digital or something more?

Blog: Mobile — Edge of Digital or something more?


Courtesy of Pixabay

Mobile has been a gamechanger in computing no less than PC decades ago. Perhaps even bigger, because it happen already deeply in the era of the Internet and it also evolved as a leading platform much faster than any computing device humans have used before. Some might argue that mobile devices also evolved itself over at least a couple of decades, but I would rather split this time between pre-touchscreen and touch screen, since only the advent of the same made them truly useful and in many ways even superior to keyboard-based devices in terms of ease and quality of interaction between the user and the machine.

But to truly understand the importance of the Mobile, we need to put it in the context of the recent history of Digital, WEB and the Internet, basically.

Indie years

iPhone has landed the mobile space with great expectations. Apple has been a battle-proven market leader in Digital Music distribution and reproduction. But when the iPhone entered the market, the reaction of the market was somewhat slow. Sales data were not impressive and to be honest, such a superior device from a hardware point of view, could bring more than just an old concept of packing it with few vendor pre-built apps, regardless of how good they were.

This has all changed only a year after with the introduction of iTunes App Store. Suddenly a cheap global channel has open for everyone! Thousands of indie devs have grabbed the chance and starting selling simple utility apps and games, which made few early adopters very successful and turning into relatively big businesses.

Those were the Indie Years. Mobile devices were super cool, but not perceived as anything Enterprise would adopt. This has changed in some ways with the release of iPad, but at first mostly inside Enterprise and B2B ecosystems.

Responsive Design and Mobile First

High-speed communications, bigger screen, pocket size, price… all this and some more has created a surge in internet usage on mobile devices and the WEB pages needed to adapt to demand.

Some old concept finally had to be shelved and challenges of Responsive Design trying to deliver the same content with constraints of size, several screens and different navigational approaches, lead to so-called Mobile First design. A big role was played here by native apps, not only because of new navigational techniques but also in the form of higher user expectations on the mobile WEB, expecting to match the same from the Native.

Enterprise overtakes the ground

Mobile Internet penetration hasn’t left the Enterprise cold. Not only the WEB was reborn as Mobile, but also B2C got a completely new playground!

B2C services suddenly got steep traction in delivery through the Native Apps, which have through their intuitive interface and ease of use spread penetration of digital B2C service far beyond the desktop and also mobile WEB pages. Mobile First got the additional new meaning, new ground: Native Client first, Desktop client fourth!

Almost in the record time, the Native Apps became primary sources of revenue. Their additional device capabilities, like Geolocating, easy to use high quality Cameras, accessibility from everywhere due to the increase of cellular data speed, have made them superior to any other client platform.

Indie was dead, long live the Enterprise!

Data on device — Native vs. Mobile WEB

Enterprise saw Native Mobile as a better client, no more. No special adaptions of the Backend were done at first. Soon the requirements coming from UX changed that and Middleware APIs were developed to cushion the performance and overall bad user experience hit caused but outdated and distributed backend APIs. Unfortunately, most of the industry is still stuck on this level and new concepts, like HTTP/2, have barely entered the stage…

But one of the biggest hits and misconceptions by Enterprise about the Mobile was understanding of the device and its capabilities. Devices are still understood and treated as sleek and fancy internet clients, nothing more. That is perhaps also one of the reasons, why we still keep reading articles peddling complete nonsense about “native is dead” and similar. Unfortunately, this shows a level of understanding of devices and Mobile in general by (too) many members of the Digital community.

A consequence of being just internet client came with a price, which is mostly paid by a user. Apps are usually slow and poorly responsive because they are being slowed down by network latency and processing of data in the cloud. Such a situation can be somewhat improved by improving APIs and middleware, but great dependency on processing in the cloud cannot deliver significant improvements in overall user experience.

But there’s the change about, coming from 3 main sources:

  • the computing power of devices keeps growing and is staggering
  • data storage on devices is already huge and keeps growing as well
  • Artificial Intelligence (AI) has taken the main stage (more about that later)

Data rulz — The Big Brother!

2+ decades of WEB and a decade of Mobile primarily through B2C have created an unprecedented amount of data. Not only social networks, every Enterprise have collected them in plenty, mostly for good, sometimes also for at least unfair if not foul purposes.

Advancements in processing powers and software support in the area of Machine Learning (ML) have given data the value, that puts the data into the centre of the business, making it the biggest asset of them all! That has inevitably lead to big concerns about the privacy, mostly justified…

But… having the data and being able to learn from it (ML), can create a whole new world of use cases for users (AI). Big Cloud data centres seem to be an ideal environment to perform machine learning, greatly benefiting from sampling from large amounts of similar data created by large amounts of the user. The situation is ideal for all kinds of algorithms of ML.

When the models are created and accepted they are being constantly improved by learning from new data, also being reinforced by user actions. But they serve the specific purpose and this is classifying new incoming data to give power to systems to make decisions based on it. Traditionally the data is analysed and classified in the Cloud as well. After all, this is the same place, where models are being re-learned and improved. Makes sense, doesn’t it?

sWell, not always. Actually, not most of the time. Why? For many reasons, but before even mentioning them, or at least the most important ones, we could summarize them into one alone:

Data is mostly first captured on the device and before being classified, it needs to be transferred to the cloud and then the results back to the device. Doesn’t seem like super efficient, does it?

Why AI on Device?

A hint comes from the cue above. But that’s only a tip of an iceberg. Reasons go much deeper than that.

Performance related reasons

  • The action happens on the device, not remote, sometimes instantly
  • No network latency, fewer energy resources spent — „Green AI”
  • Better use of strong Device CPU & GPU resources whilst saving Cloud resources
  • Less dependant on a good network connection performs better for user
  • Differential AI — accessing server resources only for deltas delivering superior experience through a hybrid approach

AI makes a difference in general, but sometimes only if it follows immediately the action it captured. Network latency prevents that at all, not only decreases the quality of user experience. Also, the devices are very, very strong today. They can handle unprecedented about of data. Actually, many devices will outperform servers in the could, because they have high performant GPU (Graphics Processing Unit), which have due to the graphics requirements specially built circuit logic for high performant matrix operations, which is ideal for AI as well!

Other important points (Safety, Privacy)

  • Processing data on the device, works with end-to-end encryption, too
  • Unencrypted data never has to leave the device
  • Enhanced network security
  • Checking integrity of data on external resources on-the-fly, keeping the users safe
  • Less legal implications (Permissions, GDPR)

Time to market

  • With less dependency on Cloud APIs and infrastructure, many AI features can be delivered through Apps only, or at least with less effort needed on the Backend

Years ago Apple made an important decision with their Apple ID and PKI infrastructure, to provide end-to-end encryption to protect the privacy of their users. This has surely created some drawbacks in a way, how they could provide good AI capabilities to some of their key apps. However, time was obviously on their side and as soon as the devices became more powerful and other algorithms improved, they were able to improve the level of their AI capabilities and whilst playing quite a good catch-up game with their biggest competitors, they deliver this by a very high level of privacy. Quite unique in the industry.

Apple has also made a step further and offer to developer a full-featured ML platform, called CoreML. It’s very adaptive and it can use models from almost everywhere. To put a cherry on the cake, they added 2 prebuilt ML modules, one for Computer Vision and one for Natural Language Processing (NLP), thus offering already very very powerful out-of-the-box ML features to developers without machine learning skills.

But not only Apple. For the same reasons other big companies are moving their AI to the Edge, to devices, to apps. Google, Facebook, just to mention a few.

The Future

Future is something, what is not easy to predict in a fast-changing environment, but there are some things I dare to say I’m pretty sure about:

  • AI, as anywhere else, will dominate the world of Mobile in the next decade, it will be the most important link between user and IoT
  • AI on the device will grow further, catch up Cloud AI and later took over the majority of tasks
  • ML on the device will overcome current limitations and eliminate performance and resource-related fears and at least in some tasks provide a useful Reinforced Machine Learning capabilities and reduce dependency on the cloud even further.

As said before, it’s very hard to predict the future, but if I would try to answer the question from the title, I would say the following:

Mobile will not be the Edge only. Mobile will share the tasks of processing the data and decision making with the Cloud and in some cases take the main role with commodity of device sensors and privacy issues as the main drivers. Logic from ML models will be used on both sides, with Cloud seeing the biggest change in user privacy, not only Data Protection. This will further push the development of ML to levels of using encrypted data, not only saving the day with Differential Privacy.

Source: Artificial Intelligence on Medium

Leave a Reply

Your email address will not be published. Required fields are marked *

Back To Top
a

Display your work in a bold & confident manner. Sometimes it’s easy for your creativity to stand out from the crowd.

Social