Apple unveils machine learning platform while Samsung turns AI focus on IoT

It seems longer than 10 years since Apple’s first iPhone ushered in the smartphone era, and now that field is at a cross roads again as markets approach saturation and margins are squeezed. AI and machine learning now figure increasingly in the strategies of all the players, especially the big three of Apple, Samsung and Huawei, which have all announced major investments and commitments recently.

Each faces different challenges, with Apple seeking to maintain its phenomenal record for profitability at this time when the others are squabbling at the lower end of the market, as reflected in smaller margins. Depending on who is counting, Samsung has held onto number one position measured by volumes with around 22%, followed by Apple on 18% and Huawei consistently placed at 11%. But Apple has 50% of revenues and for profits has 60%, compared with 25% for Samsung and just 6% for Huawei, which has cut margins to the bone in the cut throat Chinese market.

All three face slightly different challenges which they hope AI will help address. For Apple profitability is falling fast from the exalted heights of 87% in 2016 as it is forced to compete lower down the market, while Huawei is seeking ways of adding value. Samsung is challenged by a decline in smartphone sales setting in this year contributing to Q2 2018 revenues being down 4% on the same period a year earlier in a rare reversal.

The abject failure of Samsung’s Bixby voice assistant, first introduced in its Galaxy S8 early 2017, launched via a dedicated button on the handset, may have contributed to the decline because it certainly led to many negative reviews. That combined with pressure from users led to an embarrassing climbdown in September 2017, when Samsung allowed the button to be disabled so that users could enjoy the much more popular Google Assistant. This came after earlier blocking by Samsung of third party apps using that button.

Samsung was clearly going to make a comeback in AI on the smartphone but might have been well advised to drop the tarnished Bixby branding. It has not done that and instead come out with all guns blazing promoting a revamped and expanded version of Bixby as part of a $22 billion research drive in AI. But it differs from its rivals in making the primary focus 5G and the IoT, with the smartphone just an admittedly important device to be connected.

This reflects Samsung’s much stronger presence than its rivals in the consumer appliance field, with fridges and washing machines and others, as well as its growing presence in system-on-chips and other components for connected cars. The company has pledged to install some sort of AI capability in every connected device it sells, presumably including voice control in appliances. Some of this will be channeled through its Creative Lab (C-Lab) initiative, unveiled at CES 2018, which has already yielded a smart phone app called Aurora, showing a 3D character-assistant users can supposedly interact with emotionally.

Samsung is hoping that AI-related capabilities will boost its whole constellation of connected devices and appliances while also at last providing some more compelling capabilities on its smartphones.

Meanwhile Apple, whose Siri assistant has in any case been much more successful than Samsung’s Bixby, has launched Create ML – billed as a developer and training platform for training machine learning models, and then building apps. Here Apple is playing catch up, because Facebook, Amazon, Microsoft, and Google already have platforms to help train machine learning models, with the latter’s TensorFlow most popular.

Nonetheless, Apple claims, as it has done before in other contexts, to be first out with a platform that is so easy to use almost anyone can train a machine learning algorithm just through simple drag and drop operations. In a recent developer session, Apple showed how Create ML could quickly teach software to detect whether online comments were happy or angry, or to predict the quality of wine from measured characteristics including strength, acidity and sugar content. The platform is based on Apple’s Swift programming language, introduced in 2014 and feted by some developers for its ease of use.

Apple claimed its rival platforms could not be accommodated so easily into an app developer’s regular workflow and it is certainly true these others have introduced or been working on slimmed down versions such as Google’s TensorFlow Lite lightweight variant, although that is pitched more at embedded devices.

However, these rivals would argue that Create ML in its current form will be confined to relatively simple and superficial consumer-facing use cases and will be incapable of generating more complex and unique machine learning applications that would make a difference in the smart phone market. These need to be built from scratch with a lot of tuning by experienced data scientists, according to this argument which has certainly held for compelling applications of ML so far.

Another charge against Apple is that Create ML is confined to its own devices, which of course has long been the case for its earlier software, since the company’s focus is on selling more smartphones at as high a profit as possible. But that does discourage some developers who want to create apps for emerging fields such as healthcare diagnostics with universal appeal, rather than having to do the job again from scratch on Android using a radically different platform.

Then Huawei, more like Apple than Samsung, has focused its AI efforts and certainly branding on the smart phone itself. That was evident at the global launch of its flagship Honor 10 mobile phone in May 2018, under the slogan “beauty in AI”. There are two clear applications of ML here, both associated with the camera, which is pitched as its main differentiator.

First is Huawei’s image stabilization technology, designed to eliminate the curse of camera shake, and then related to that is object enhancement, which can be likened to automated photoshop, where images inside the picture are sharpened with edges standing out. The company hopes such features will allow prices to hold up slightly better so that the gap between profits and volumes can be narrowed – as the camera is such an important part of the phone’s design and consumer preferences, these days.

If there is a common trend it is towards the smartphone being the hub of a collected ecosystem of devices embracing cars as well and AI will be the banner as well as the technological glue joining them together as data flows increasingly between them.