What is AI and how did it get into your phone?
Artificial Intelligence (AI) is really an umbrella concept, comprising subsets of specific technologies that you’ve probably heard bandied around, including ‘Neural Networks’ and ‘Machine Learning’. Essentially, AI boils down to ‘fuzzy pattern matching,’ performed by computers, which rely on huge datasets, and enormous amounts of processing power to get their results.
To give you an idea of the scale of the calculations involved in these tasks, some estimates suggest that AI algorithms could potentially require up to 10 million data points, to match human level performance at any specific task.
AI is already ubiquitous, across the economy, and behind our interfaces to the internet. McKinsey says the technology is already part of 69% of our digital interactions. And its contribution has only just begun.
AI can be used in a literally staggering number of places to provide value for businesses, from targeted recommendations and advertisements, both online and off, to filtering out fake news and identifying Twitter Bots pretending to be people. It’s employed in optimizing energy uses across data centers and is the guts of every Google / Bing Search you perform. It offers huge efficiencies to logistics and freight companies by optimizing delivery routes. It assists in lowering servicing costs and downtimes by providing predictive maintenance, on everything from jet engines to electric vehicles. It’s the technology behind Amazons’ personalized recommendations and can even be used by Police forces to identifying faces and license plates in real time.
The Smartphone AI conundrum
Computer processors, server farms and cloud-based data centers are using more energy than ever, performing the massive quantity of calculations that are required to. According to Qualcomm by 2025, (link above) data centers could use one-fifth of the entire world’s energy supply.
As a result of the datasets and calculations involved, AI is incredibly ‘expensive’ in computing terms. It uses a lot of energy and many AI tasks have to be done quickly. Resource intensive applications are the opposite of what phone makers want. Samsung, LG, Huawei, and Apple want to provide their user’s better battery lives and more efficiency. That means the growth in the use and usefulness of AI presents a unique challenge to those who design and manufacture mobile phones.
Source: Qualcomm 2017
A recent example of AI in smartphones – designed to address these challenges
Don’t underestimate the importance of getting the balance of improved on-device experience, as a result of AI tools, and reduced battery performance. Take Apple’s Bionic processor in the iPhone X Series, announced this month, for example. Apple’s XS, XS Max devices contained no less than 3 different types of the microprocessor, the range designed to maximize computing power and minimize battery drain on this year’s devices. The iPhone XS and MX Max contain:
- An A12 CPU (Central Processing Unit – the main brain of your phone): The first 7nm CPU to be sold anywhere in the world, designed with 6 cores, 4 of which are targeted at powerful processing tasks, and 2 of which are designed to operate efficiently when little strain is on the phone – for example, when it’s not being used.
- A separate GPU (Graphical processing unit): GPUs are designed to efficiently conduct the 3D calculations required to show accurate images on your phone. This processor too has multiple cores.
- A ‘Bionic’ chip: This is chip Apple has started including specifically to assist with AI calculations.
All of this focus, all of this hardware, is designed to provide the best balance possible between a fast, enjoyable user experience on this year’s iPhones, without compromising the phone’s. Indeed Apple promises a 30 minute to 1-hour charge advantage in their new range of phones. In large part, the benefit is derived from this new processor layout.
What AI on your smartphone provides
Ai is already a mainstay of the tools we take for granted on our phones. It’s used for:
- Searching for pictures on your phone by the wording in them or: Many phones will now collate pictures of an event or day into a ‘Story’ and provide you with an entertaining summary. Often, they can be shared with friends.
- ‘Recognize’ things in human ways: AI on your phone can identify individuals from their pictures and even determine the emotional state associated with particular facial expressions.
- It’s the technology behind Siri and Google’s ‘Personal Assistant’: Interacting with an on-device chatbot or personal assistant, especially when you do it with your voice, requires AI smarts to figure out your intent and recognize your ‘utterance’ (AI talk for what you said.)
- AI is what transposes your spoken messages: Typing dictated messages even providing the human voice that reads the text messages to you are also AI tasks.
Unfortunately, Phones are essentially advanced snooping devices, filled with sensors. As a result of the sheer number of measurement components and constant mobile internet connectivity, AI also presents the risk of hacking to users, too. Recent (thankfully lab-based) research suggests that AI algorithms can be used to Identify your ATM PIN from the way your phone moves in your pocket when you enter it. Other, similar, work shows that researchers could uniquely identify a user, within 4 hours, from the way their finger scrolls across a smartphone screen.
Smartphones are Edge Computing devices and they’re about to be turbo-charged with 5G
Edge computing is developing as a way to gather, work on and filter the data needed to conduct an AI action, and send back to the cloud a parboiled answer. Edge computing performs the processing of data gathered from connected devices, close to the source of that data.
At a principal level, smartphones are the archetype and among the most advanced Edge computing devices used in the world. Any mobile device, operating in the real world may not always be connected to the mobile internet. That means, from time to time, they are ‘operating under their own recognizance.’ Embedding AI chips in smartphones make that Edge Computing as inexpensive in both time and battery usage, as is humanly possible.
It’s flagship devices which contain specialized AI chips now, but the technology will filter down to cheaper devices in time. With little to pick between new device releases, dedicated AI chips might be the only reason we have to upgrade our phones, in the future. 5G network connections, between 10 and 100 times faster than the current 4G standard will come to market as early as 2020 and will provide even greater bandwidth. AI is already central to how we experience our phones and it’s only going to get smarter when Edge Computing is connected to more powerful AI services in the cloud, at 5G speeds.