As 2019 comes to an end, Synced takes a look back at the year’s bumper crop of AI-powered products, spotlighting 10 new apps that we believe present brilliant ideas, innovative design and unique functionalities.
Google Recorder – Transcribe recordings automatically
At its Made by Google 2019 event the company introduced the new “Recorder” voice recorder app for Android devices. The app uses AI-powered speech recognition to automatically transcribe recordings of lectures, meetings, interviews, etc. Google recently detailed how they leveraged recent developments in on-device machine learning to power the Recorder app.
Seeing AI – Microsoft helps the blind explore photos by touch
Since 2017, Microsoft’s Seeing AI has been using AI technology to help blind and vision-impaired persons navigate through their everyday lives. This year, the company released a Seeing AI update that enables users to explore photos by touch. Microsoft leveraged technology from its Azure Cognitive Services to develop the new feature. A tap of the finger on a touchscreen image produces audio descriptions of objects in the image and the spatial relationship between them.
Replika – Chatbot companion for mental health
This Y Combinator-backed AI firm trained a compassionate companion chatbot to learn from users. Moreover, the bot speaks in an appropriately soothing voice and humanlike manner. One of Replika’s research goals is to develop tools that can help users deal with problems such as loneliness, and mental health issues. The new voice recognition feature enables the chatbot to call users on the phone and talk about feelings.
AI Dungeon – Text adventure game with AI
AI Dungeon is a text adventure game where users explore an infinitely AI-generated text world. Unlike than other adventure games with actions limited to plotlines designed by developers, AI Dungeon enables endless story possibles. Says AI dungeon creator Nick Walton: “Anything you can express in language can be your action and the AI dungeon master will decide how the world responds to your actions.”
AiCure – Monitoring medication-taking
New York-based AiCure created the app to help users monitor their medication-taking. The elderly for example can sometimes duplicate or take unnecessary medications, or forget to take their medications at the right time, which can adversely affect health. AiCure uses a smartphone camera and AI algorithm to monitor users and issue timely reminders if meds are missed or not taken properly.
Whisk – Precision recipes using AI
In March the Samsung NEXT Product group acquired the smart food platform Whisk, to advance meal planning and recipe finding services. Providing more than 500 million recipe interactions every month, Whisk’s Food AI leverages millions of data points covering almost any recipe ingredient in the world, its relationship to other ingredients, and its nutritional value.
Fujitsu – Disaster evacuation app for smartphones
Japanese electronics giant Fujitsu has developed an evacuation app for smartphones designed to help vulnerable individuals make informed and timely evacuation decisions in case of fire or flood. In the 2011 Tohoku earthquake and tsunami disaster struck, many local residents suffered from evacuation delays made worse by lack of information.
The app computes and displays real-time evacuation strategies that show users’ current location and nearby safe evacuation locations and routes. The app also displays the probability of flooding in nearby locations as predicted by an AI algorithm, along with the number of people who have already evacuated to a particular shelter.
PlantVillage Nuru – Helping farmers diagnose crop disease without an internet connection
Researchers at Penn State University developed the app, which uses Google’s TensorFlow machine learning tool and a database of images collected by crop disease experts around the world to help farmers diagnose crop disease. The free app was developed in cooperation with the International Institute of Tropical Agriculture and the United Nations Food and Agricultural Organization.
Predicting near-term crop yields for farmers in Africa could help protect popular crops from threats posed for example by climate change. The app is based on research comparing ML model and human expert accuracy and extension work and is frequently updated. The app also allows for joint model where both AI and human experts collaborate on image analysis through a cloud system.
ChimpFace – Facial recognition combats online wildlife trafficking
The Conservation X Project’s “ChimpFace” identifies and marks chimpanzees in photos on social media and e-commerce websites to help detect potential online wildlife trafficking. The model reduces the time spent by humans monitoring online wildlife trafficking activities and is an efficient and scalable tool for conservation organizations and law enforcement alike.
Google Play Store – Machine learning and personalised apps
Last but not least is the Google Play Store itself. Researchers at Alphabet’s DeepMind have detailed a collaboration with Google to personalized users’ app-discovery experience. For example, to improve how Play Store’s recommendation system learns users’ preferences, DeepMind replaced an LSTM (Long Short-Term Memory) model with a Transformer model, which is better equipped for sequence-to-sequence prediction and has yielded strong results in NLP.
Journalist: Fangyu Cai | Editor: Michael Sarazen