AI and Your Smartphone Trivia Quiz

Answer: Improved privacy

Generally, the more AI functionality added to a smartphone, the more demand on the battery resulting in less time between charges. The obvious solution is to use cloud-based processing accessed via a cellphone or WIFI networks, although higher network usage will also impact on battery life.

On-device AI however offers several advantages: increased privacy (facial recognition data stays on the phone), faster performance (using networks introduces latency), added offline functionality (not everywhere has a good network signal) and improved security (keeping financial and health data as local as possible gives hackers less opportunity). Using specialised on-device Neural Processing Units (NPUs) and optimising AI models are some of the techniques used to keep the data processing on the smartphone. As battery life has improved, more on-device AI has become possible without compromising performance.

Answer: Natural Language Processing

In this case NLP stands for Natural Language Processing and acts as the interface between you and the smartphone, whether you are speaking to it or typing text. These include speech recognition (if you were dictating a document), text-to-speech (where the phone gives you GPS navigation prompts), performing autocorrect and predictive text functions, intent recognition (understanding system commands) and named entity recognition (for example, so that phone calls can be placed to the correct person).

Other NLP techniques more typically found in apps allow real-time translation between languages, understanding the tone or meaning in messages, and powering specialised conversational chatbots.

Answer: Phoning using just voice commands

NLP and OS integration typically allow the user to make calls, send texts, open apps, and adjust settings. All are standard embedded functions or apps found on most, if not all, smartphones. Generally the NLP acts as the interface layer between the phone and user. Other AI or technology layers handle the particular task required.

For e-commerce, a payment system and a supplier API (application programming interface) would be required in addition to NLP. For the navigation & location services, the layers are NLP, GPS (Global Positioning System), maps API and geocoding (takes a street address and converts it to geographical coordinates for the map). For media & entertainment, you will typically link to a media app. The media app may come with its own recommendation system or the smartphone AI could make suggestions based on previous selections.

Answer: Yes

The battery lifespan can be extended by managing the charging process. If charging at night, and the smartphone knows when you habitually start using the phone in the morning, it can delay charging past 80% until just before you usually use the phone. This reduces the time the battery spends at full charge which in turn reduces the chemical ageing process that lithium batteries are prone to.

Other optimisation strategies revolve around things like knowing which apps get used frequently (to keep 'warm' or to hibernate), which apps have heavy processor needs or not (to allocate the right processor resources), and delaying synching when the battery charge status is low.

Answer: Classification of scene type

The image seen by the camera is assessed against a pattern-recognition library and classified into a particular category. The smartphone will normally flag this up, for example, if a picture of text is being framed. Based on this, the camera settings will be automatically adjusted so that the resulting photo is optimised for that type of scene.

This might, for example, involve boosting colour saturation and sharpening edges for landscape scenes. It is all done more or less instantly as the shot is framed.

Answer: By blurring the background

This is the smartphone equivalent of the bokeh effect, a technique used to produce an aesthetic quality of blur to out-of-focus areas of a photograph and thus place emphasis on the subject. Dual-camera smartphones might achieve this by capturing two simultaneous photos, one blurred, and using software to combine them. Single-lens smartphones can achieve the same effect using computational photography where multiple photos are taken with different settings and software is then used to produce the desired effect.

Computational photography also makes achieving good photos in low-light levels or in high-contrast situations (such as strong sunlight with deep shade) possible with the smartphone camera.

Answer: Facial recognition

Facial recognition is the biometric method most commonly enhanced by AI for spoof detection in smartphones, using techniques such as liveness detection to differentiate between real users and spoofed examples. As you might expect, liveness detection is about verifying that the biometric data presented comes from a live person rather than a fake image or video.

Two methods are used: active liveness detection (where the user to perform specific actions like blinking or smiling) and passive liveness detection (which analyzes subtle features like skin texture and light reflection without requiring user interaction).

Answer: Adjusts facial feature maps to account for ageing, lighting, and angles

When you unlock your phone with your face, AI algorithms analyze dozens of points on your face, measuring distances between features to create a unique facial map. This process is adaptive so the more you use this feature, the better the smartphone gets at recognising you even if your appearance changes over time. Unlike early face unlock systems that could be fooled by photos, modern AI-based recognition includes liveness detection. Newer systems might include eye movement and voice for layered security. Having a beard, hairstyle changes and wearing glasses can be accommodated by the AI.

Answer: By using machine learning to detect partial or distorted prints

Fingerprint scanning has also evolved with AI. Traditional fingerprint sensors compared patterns linearly, but AI-enhanced biometric systems use complex pattern recognition, capable of learning subtle changes over time - like slight injuries or skin texture variations.

In addition to biometrics, AI monitors for suspicious behavior. If someone tries to hack into your phone through unusual login attempts or odd app behaviors, this can trigger security protocols which lock the device or alert the user.

Answer: Near a school playground

Sudden, irregular sounds typical of school playgrounds are harder to deal with than constant, regular noise backgrounds. Unlike static filters, AI can be trained to recognise human voice features so that background sounds such as typing, dog barking or traffic can be suppressed. Adaptive filtering is the key to this.

Smartphones are equipped with more than one microphone, some of which are used for active noise cancellation. They pick up background noise, record it, invert it and add it to the voice mic's output (thus cancelling the background noise picked up by the voice mic) before, for example, a caller hears you or your instruction goes to the virtual assistant. The multi-microphone combination also helps with working out where a voice is coming from, so voices picked up from a TV can be safely treated as background noise and filtered out.