AbstractA mobile deviceis a portable computing device that performs many functions of a computer witha mobile operating system and integrated cellular network connection forcommunication. Artificial intelligence (AI) is the ability of a computer systemto perform tasks that requires human intelligence like decision-making, speechrecognition, and translation between languages. Artificial intelligence isbeing used in automotive industry for self-driving cars, video-games, healthcare, etc. With the advancement in technology and innovation in smartphonemanufacturers, artificial intelligence made its way into mobile devices. AI in mobiledevices learns about the user preferences over time by tracking the activities.The data collected by AI is stored in the device which keeps it personal andunder the user’s control. Some of the apps share this information anonymouslyto their servers to enhance their AI.
Although AI is in nascent stage, there isa lot of controversy about its advancement. Experts believe that AI will soonturn smartphones into intelligent entities that will know how we feel and whatour emotions are. Research is taking place on the said AI technology which willknow the mood of every person and how it is likely to change.
Such advancementsin AI will give mobile devices the power to learn our emotions and to predictthem. AI has made everything from virtual assistant to face recognitionpossible on a smartphone. Tech giants, Google and Apple have already integratedartificial intelligence into their maps; Apple Map or Google Maps as they cannow predict and make suggestions on where you might want to go. Another popularapplication of AI that comes with modern smartphones today is the AI personalassistant. To name a few, we have Apple’s Siri, Microsoft’s Cortana, GoogleNow, and Huawei’s upcoming Kirin 970. The continuous improvements to the mobiledevice AI assistants made them more ‘human’ like in various aspects such as theway they talk, behave, and act.
Further innovations in the AI technology would significantlyaffect the way these virtual assistants would work. Besides all theadvancements and advantages of AI in mobile devices, it also poses some seriousthreats. The vulnerability of mobile devices to electronic spying and hackingadds up to the existing threats in AI. This paper provides a study on “Thepotential implications, benefits, and challenges of integrating AI technologyinto mobile devices”? Keywords: ArtificialIntelligence, Smartphones, Technological Advancement, Human Intelligence,Virtual Assistant1.IntroductionFrom sciencefiction to reality, artificial intelligence is now an everyday term which onceenjoyed a superior status. John McCarthy is the Father of ArtificialIntelligence (Rajaraman, 2014). He coined the termArtificial Intelligence in the year 1955 that describes computer programs whichdisplay intelligence, that is, these programs have the ability to perform taskswhich when performed by humans potentially requires intelligence (Rajaraman, 2014).
He invented LISP, a programminglanguage designed specifically to solve the issues in Artificial Intelligence (Rajaraman, 2014). A rise in artificial intelligencewas observed from 1957 to 1974 (Anyoha, 2017). It was the time whencomputers became faster, cheaper and easily accessible (Anyoha, 2017).
Also, there was an improvement in machinelearning algorithms and people started understanding the best fit algorithms totheir problems (Anyoha, 2017). Machine learning is anincreasingly powerful tool which can be applied to wide variety of areasranging from object recognition, language translation, and more (Thorat & Smilkov, 2017). This makes use of learningalgorithms that helps computer systems to learn new tasks instead ofprogramming them continuously to perform tasks (Society, 2017). Figure 1: Artificial IntelligenceEvolution (Anyoha, 2017)The above figureshows the timeline of artificial intelligence starting with whether machinescan think in 1950’s to inventing the first speech recognition software in2000’s (Anyoha, 2017).
A trend incomputer technology which is observed over the past decades was that thedevices turning portable and personal (Malaka, n.d.). Mobile devices are designedin such a way that they are small enough to be carried around with computingcapabilities of a personal computer or desktop (Tan & Lindberg, 2008) (Wikipedia). With significantprogress in technology in the past few years, mobile devices now have becomemultifunctional devices that are capable of handling wide range of applications(Tan & Lindberg, 2008). For example, besides makingphone calls and sending messages smartphones are capable of connecting to theinternet enabling sending of e-mails and web browsing (Tan & Lindberg, 2008). For the success of mobilesystems, the design of user interface and software play an important role thatallows access to personal information and provides flexibility in handling changesaccording to user requirements (Krüger & Malaka, 2004). There are a lot of difficultiesthat needs to be solved before building useful infrastructures for mobile users(Krüger & Malaka, 2004).
This is where artificialintelligence (AI) comes into picture and helps in solving many of the userdifficulties (Krüger & Malaka, 2004). AI technology is madeavailable in mobile phones for a while now, providing voice assistant featureslike Siri, Cortana, or Google Assistant (Oliva, 2017). Additional AI applicationsin mobile devices include: Maps that compute the optimal driving route anddistance to destination, song suggestions, friend recommendations on variousapplications, job recommendations based on previous searches, productrecommendations, all these are a result of AI in mobile devices (Deloitte, 2016). This paperprovides information that helps in understanding the implications, benefits,and challenges on AI in mobile devices.1.1 Research Question To understandthe potential implications, benefits, and challenges of integrating AItechnology into mobile devices.2.
Literature ReviewArtificialintelligence is firmly rooted into our day-to-day lives (Deloitte, 2016). Though AI technologies havebeen in existence for several decades, the explosion of massive amounts ofdata, which is the raw material for AI that led to its rapid advancement (Marr, 2017). Many voice-controlled intelligent personalassistants, such as Cortana by Microsoft, Siri by Apple, Google Assistant byGoogle, Bixby by Samsung, are progressing to be a part of users’ daily lives, largelyon mobile devices (Kiseleva et al., 2016). There is a significant differencewhen information is accessed by the intelligent assistant in contrast totraditional web search (Kiseleva et al., 2016).
Figure 2: Interaction with voiceassistant: Cortana (Kiseleva et al., 2016)Figure 2 showstwo examples of tasks performed by the assistants (Kiseleva et al., 2016). These intelligent assistantshelp users with communication, information, and time management (Azvine, Dijan, Tsui, & Wobcke, n.
d.). Each of these assistants aredesigned to have a model of their user and a learning model to betterunderstand their users and to provide more personalized services (Azvine et al.
, n.d.). Users communicate with theirassistants in order to control their mobile devices such as making a phonecall, setting up reminders and alarms, or to manage their calendars (Kiseleva et al., 2016).
These interactions are madepossible with the help of automatic speech recognition (Kiseleva et al., 2016). Another important applicationof artificial intelligence in mobile devices is its usage in Global PositioningSystem. GPS navigation systems make use of stored map information for selectingthe best route (Duffany, 2010). GPS system uses measures GPS location, date,and time information in determining the best possible route (Duffany, 2010). Learning based artificialintelligence provides GPS system with the ability to recognize traffic, trafficlights, stop signs, individual driving habits, weather, direction of travel,speed limits, day of the week and time of the year between the source anddestination points (Duffany, 2010). Further, recommender systems(RS) in various mobile applications use artificial intelligence (AI) methods toprovide recommendations to the users (Portugal, Alencar, & Cowan, n.d.
). Artificial intelligence alsoaims at improving the level of safety from malware attacks in android operatingsystem mobile devices (Lopez, Cadavid, & Garcia, 2016). This is achieved by machinelearning classifiers where the algorithms are trained with known features inorder to predict the classification of unknown features (Amos, Turner, & White, 2013) . Another newest feature added tomobile devices by integrating AI is Google Lens launched by Google in itsmobile phones Pixel and Pixel 2 (Statt, 2017). This makes use of thetechnology computer vision (Statt, 2017). 2.
1Artificial Intelligence Benefits and Technologies:Artificialintelligence tends towards creating a personalized user experience (Wertz, 2017). It analyzes enormous datasets efficiently than a human being (Wertz, 2017). It can rapidly identifypatterns of information, such as purchase trends, credit checks and othercommon threads which are analyzed every day to cater to a single customer (Wertz, 2017). For example, the AI-poweredfeature of Now Playing in Google Pixel 2, with a database of 100,000 songs thatare updated weekly, will tell the user the song playing in the background byautomatically sending a notification onto the home screen (Oliva, 2017). Another example is theGoogle Translator that understands different languages and translates them tothe user language (Oliva, 2017). The most important benefitof AI is that it makes the mobile devices better understand the way user usesit and will provide more relevant features and applications (Oliva, 2017).
To achieve these benefitsfollowing technologies are incorporated with AI.Forrester’sTechRadar methodology is used in identifying and analyzing the current and futuretrends of different AI technologies (Forrester, 2017). Figure 3 shows the Forresterreport on artificial intelligence technologies. Figure 3: Artificial IntelligenceTechnologies (Forrester, 2017)There is asignificant rise in investment and adoption of artificial intelligence by manytech giants (Press, 2017). The most widely used AItechnologies are:· Text Analytics and Natural Language Processing:The ability of a computer to understand and communicate with humans in theirnatural language (Weischedel et al., n.
d.). NLP makes use of textanalytics to understand the sentence structure, meaning, sentiment and intentthrough machine learning methods (Press, 2017). NLP is used in number ofdisciplines such as computer and information science, linguistics, mathematics,electrical and electronic engineering, artificial intelligence and robotics (Chowdhury, 2003).· Speech Recognition: This is an ability of thecomputer to map an acoustic speech signal to text (Forrester, 2017). · Facial Recognition: The ability to analyzeimages and videos of human facial expressions to identify users and to detecttheir emotional responses and engagement levels (Forrester, 2017).· Machine Learning Platforms: Making availabledifferent development and training toolkits, models, algorithms, computingpower to applications and machines (Press, 2017).
· Deep Learning Platforms: This makes use ofartificial neural networks that provide computers with the capability to learnand improve on their own (Parloff, 2016). This is primarily used inpattern recognition and classification applications (Press, 2017). For example, recognizing themake and color of a car, identifying the landmarks in picture (Triggs, 2017).· Biometrics: This enables the interactivitybetween humans and machines in the form of image and touch recognition, speechand body language (Press, 2017). These are majorly used for authenticating oridentifying the user (PRNewswire, 2016).
2.2 Challenges of AI in mobile devicesThough there aremultiple benefits of integrating AI in mobile devices there are few challenges associatedwith it. They are:· With the increase in Neural Networks, MachineLearning, and Heterogeneous Computing there are emerging use cases forsmartphone users spread across a wide range of fields (Triggs, 2017). These technologies improveuser experiences by offering them with enhanced audio, image and voiceprocessing, language processing and improved database search speed, amongothers (Triggs, 2017). Due to limited processing capabilities ofsmartphones these use cases cannot be easily obtained (Mao, Zhang, Song, & Letaief, 2016).
Hence it is harder toimplement artificial intelligence on a mobile platform (Zhang, Gu?sgen, & Yeap, 2004).· The artificial intelligence revolution involvessending information to vast data centers on the cloud server, where it isprocessed before obtaining the results (Burgess, 2017). Sending data back and forthis slowing down the processing of data and also posing serious privacy concerns(Burgess, 2017).· Another serious concern of AI is its Intuitivethinking or Judging power.
That is, artificial intelligence or machine learningworks by training the software to identify patterns or trends in the data (Nogrady, 2016). After it is trained, itperforms analysis on fresh, unseen data and derives the results (Nogrady, 2016). There is no proper reasoningon how it got to that particular result (Nogrady, 2016). Therefore, the performanceof the system depends on the data it learns from (Nogrady, 2016). Since the data that is fedto AI’s is not perfect, the results obtained are not accurate (Nogrady, 2016). There should be morescrutiny on the decision-making process of AIs (Nogrady, 2016).
2.3 Artificial Intelligence in Mobile Apps The expansion ofAI technology has led the mobile users to expect an improved user experienceand mobile app performance (Dossey, 2017). With the increase in userexpectations many companies have developed their intelligent mobileapplications (Dossey, 2017). Apps are the first steptowards introducing AI to mobile devices (Burgess, 2017). For example, Starbucksreleased its mobile app “My Starbucks Barista” that places orders for users (Dossey, 2017). Another example is of”TacoBot” released by Tacobell that recommends personalized menu suggestions basedon the user prior purchase trends (Dossey, 2017).
Few such apps are “The Roll”,this organizes thousands of phones taken with the phone by grouping similarpictures together thereby deleting the copies (Eaton, 2016). “EasilyDo” which connectsthe email accounts like Gmail to other services and apps like Facebook,LinkedIn so that it can send an alert reminding you of the scheduled meeting inthe calendar (Eaton, 2016). There is a significantgrowth in development of these smart apps as they help in accomplishing dailytasks (Dossey, 2017). Many apps are beingdeveloped with algorithms that adjust based on the user behavior (Dossey, 2017). These algorithms help tocreate more meaningful applications that can engage users (Dossey, 2017).3.Applications of AIin mobile devices:3.1 Voice controlledpersonal assistantsThere is a significant growth in the usage of voice-controlledintelligent devices, such as Microsoft’s Cortana, Google’s Google Assistant,Apple’s Siri, and Samsung’s Bixby (Kiseleva & de Rijke, 2017).
These personal assistantsassist user in performing different tasks, both at work and in their dailylives considering context as a crucial factor (Kiseleva & de Rijke, 2017). For example, they help inmaking phone calls, planning night outs, performing web search and many more (Kiseleva et al., 2016). 3.
2 Face DetectionDeep learning techniques are used to detect faces (Apple, 2017). This technique captures theface attribute features available such as age, emotion, gender, smile, andfacial hair (Microsoft, n.d.).
Example of this is theiPhone X “Face ID” feature which is used to secure the phone (Staff & Fleishman, 2017).3.3 RecommenderSystems Recommender systems are vital in mobile commerce as itbenefits users by providing highly personalized products and services (Frey et al., 2017). For example, on a social networking app, RS suggests profileswhich are similar to user, based on their interests (Portugal et al., n.d.
). Another example is ofAmazon, it is known for its personalization and recommendation, which helpscustomers find items of their interest by using these recommender systems (Linden, Smith, & York, 2003).3.4 Computer Vision SystemsComputer vision deals with the technology of imageprocessing, pattern recognition, physics, and artificial intelligence (Shapiro, 1985). An example for this is “GoogleLens”. Lens is a computer vision system that is capable of retrievinginformation of any object in real time by pointing the Pixel or Pixel 2 cameraat it (Statt, 2017). It can read text and savesinformation from business cards, saves URL from fliers, identify landmarks, canretrieve information on books, art, posters by focusing the camera on to them (Statt, 2017).
Lens also works as barcodeand QR code scanner (Statt, 2017).4 Future of AI in mobile devicesTechnology isever changing and there is a constant advancement in the capabilities providedby artificial intelligence. Besides the general applications of AI in mobiledevices mentioned above, many mobile tech giants are moving towards enhancingthe use of artificial intelligence to its fullest in their mobile devices. Toaccommodate the benefits offered by this ever growing technology on mobiledevices, a new custom AI processor thatcan handle these tasks is to be designed (Triggs, 2017). This not only improves thecomputational power but also efficiency in three main fields: size,computation, and energy (Triggs, 2017). An extra dedicated processorfor handling complex sorting algorithms will help improve the computationalpower of smartphones and enhance faster processing, from automatic imageenhancements to faster video library searches (Triggs, 2017). Major mobile vendors areworking towards implementing this custom chip on their mobile devices forimproving its computational power.
Apple, the biggest mobile phone vendor believesthat its mobile devices will be a major platform for implementation ofartificial intelligence (Reuters, 2017). Apple’s new iPhone X is billedas “the future of the smartphone” with its new enhanced neural engine with acapability to process “up to 600 billion operations per second” (Vincent, 2017). Till today, AI features thatare made available on mobile devices are powered by the cloud (Vincent, 2017). This is one of the majorshortcomings as a steady internet connection is needed to make it work and itis less secure as the personal data is sent to the main servers (Vincent, 2017). To overcome this, Apple isfocusing on doing AI on the local mobile devices (Vincent, 2017). By incorporating AIcompatible hardware in phone itself protects users’ privacy by sending lessdata off-device (Vincent, 2017). Chinese tech giant, Huaweiis building a neural processing unit on its Kirin 970 system-on-chip which hashigher processing speeds and can process image recognition 20 times faster thana regular CPU (Vincent, 2017). It is the first advancementin bringing powerful AI features to the local mobile devices (Business Wire, 2017).
On-Device AI is equippedwith capabilities that lays the foundation for understanding and assistingpeople (Business Wire, 2017). Sensors present in theon-device AI are capable of producing huge real-time, personalized data (Business Wire, 2017). Backed by powerful chipprocessing capabilities, devices will become more cognitive of user needs andwill provide highly personalized services (Business Wire, 2017). Google also is working on methodsto implement on-device AI called “federated learning” (Vincent, 2017). Further, research is goingon at Google in making google maps find parking spots on arriving atdestinations and to distinguish between people using taxis versus actuallydriving the car (Bohn, 2017).
The most important factorpeople are more concerned about is privacy and how to mange the privacysettings on their mobile devices (Bohn, 2017). Google is positive that AIcould fix this problem heuristically by making the systems more sophisticatedtowards user emotions (Bohn, 2017). With the increasingadvancements of AI in mobile devices, chip design companies ARM and Qualcommare configuring AI compatible mobile chips (Vincent, 2017). With the ever increasingcyber attacks improvements to mobile security is important (Howarth, 2016). Mobile security can bepreserved with the use of machine learning that helps the organizations inbetter interpreting of the data which leads to effective detection of securitythreats and vulnerabilities (Howarth, 2016).
Further research in AI will lead to solutionsto increasing cyberattacks (Howarth, 2016).5 Implications of AI in mobile devicesThe growth ofartificial intelligence is creating a whole new world for mobile devices. It has advanced to such an extent where itshows messages and images that correspond to user thoughts and opinions (Forbes, 2017). The ads that are recommendedto the user are tailor made specially for that user based on his prior searches(Forbes, 2017). AI enhanced chatbots areused for quick communication (Forbes, 2017).
For examples, messenger appby Facebook has enabled 11,000 chatbots that allows its users to do anything (Forbes, 2017). In search technology, voicesearches increased from ‘statistical zero’ to above 10% of global searches in2015; About 25% of windows 10 taskbar searches were made via voice as reportedby Bing, and 20% of mobile android searches are by voice which shows the newexpectations set by AI (Accenture, 2017). People interact with theirsmartphones regularly to check the weather app every morning or to book an Uberafter checking Citymapper in the night; AI based interface will learn thisbehavior and perform them automatically as a routine (Deloitte, 2016). With the growth of AI towardsuser experience, it might soon be more than an intelligent interface (Accenture, 2017). Because of its simplicity,each user engagement might become more personal, powerful and natural (Accenture, 2017). Deploying AI acrossinterfaces will open doors towards greater adoption of complicated tools (Accenture, 2017). AI tools are seamlesslyintegrated with mobile devices which made them an essential component today (Accenture, 2017).
The way developers and userssee intelligent interactions within mobile applications is taking a profoundturn with the advancement and availability of machine learning (Dossey, 2017). 6 Personal ContributionAfterresearching rigorously on artificial intelligence, I can say that AI ischanging the world at a breakneck speed. The goal of artificial intelligence isto improve the user experience and make it more personalized.
In order toachieve this, the mobile vendors like Google, Apple are collecting user informationsuch as location details, frequently used app details and how these apps areused by the user with the help of the mobile devices and sending them back totheir cloud servers. This invades the user privacy. AI requires tons of data toperform analysis which makes privacy implications bigger. There is a potentialthat a large amount of personally identifiable data is being gathered. Ifproper masking or de-identification techniques are not performed beforeperforming the analysis on user data, it can lead to leak of personalinformation. Sometimes not performing anonymization properly can also lead tostigmatizing analytics which can have a significant negative impact on a useror a group of users. Solution to protect privacy of a user can be achieved byimplementing AI on-device instead of using the cloud server as used currently.
This is still in its nascent stage and major tech giants are working towardsachieving this.With steadyincrease in chatbots in different applications for helping users to get variousproducts and services, it will have a serious impact with respect to human jobs.Bots being cheaper and quicker to develop are gaining prominence significantly.Few jobs that are already automated by bots are Customer Service agents andFast-food servers. For example, Tacobell’s Tacobot.
There is a possibility thatin the coming years these chatbots might replace human workforce.