Month: August 2017

In the movie The Terminator, directed by James Cameron and released 33 years ago, artificial intelligence or AI was already showcased. One of the scenes shows how the AI robotic killer, played by Arnold Schwarzenegger, welded the damaged parts of his body together in a motel room. A cleaner passing by noticed the smell and asked if there was a dead cat inside. A few options for the response popped up for Arnold in the movie:

“Yes”/”No”/”Please come back later”/”Piss off!”

The intelligent robot picked the last answer which best suited the situation. As a result, the cleaner went away without any suspicion. Interestingly, we are still developing such sophisticated AI.

AI has a broad meaning. From a technical perspective, it includes:

• Deep learning – learning from a large pool of data to assimilate human intelligence, such as AlphaGo in the Go world;
• Robotics – responsible for pre-determined extremely difficult or dangerous tasks, such as surgery, dismantling bombs, surveying damaged nuclear power plants, etc;
• Digital personal assistants, such as Apple’s Siri, Facebook’s M;
• Querying method – finding information from a huge database speedily and accurately, such as IBM’s Watson, which takes only 10 minutes to identify a rare leukaemia after searching through 20 million medical papers; or the Al-DR from China, which is said to diagnose lung cancer cells in 0.1 seconds through X-ray films;
• Natural language processing, such as a chatbot called Ali Xiaomi (Ali Assistant) at Alibaba’s Taobao. Ali Xiaomi handles both spoken and written queries, including providing answers to frequently asked questions and questions about specific transactions such as delivery status;
• Context-aware processing, such as when we move the mobile phone from vertical to horizontal view, and screen view is automatically rotated.

However, what are our expectations of AI? In “The AI Revolution: The Road to Superintelligence”, the author points out that there are 3 stages of development in Al:

Basic — Artificial Narrow Intelligence or Weak AI, i.e. AI specializes in a certain scope, IBM’s Deep Blue can beat the world’s invincible hand in chess, but I am afraid it’s unable to guide you to the nearby restaurants or to book a hotel room for you. The same logic applies to bomb disposal robots and the AI which identifies cancer cells within seconds.

Advanced — Artificial General Intelligence or Strong AI, i.e. the computer thinks and operates like a human being. How does a human think? I have just read a column from a connoisseur. “There are many factors affecting us in choosing a catering place, like our mood, type and taste of food, price, time, etc. The determined factors are not the same every time.” See, it is really complicated. There is a so called Turing Test. Alan Turing, a British scientist who was born over 100 years ago, said: “If a computer makes you believe that ‘it’ is human, it is artificial intelligence.”

Super Advanced – Artificial Superintelligence. Nick Bostrom, a philosopher at Oxford University, has been thinking about Al’s relationship with mankind for years, and he defines superintelligence as “an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills.”

Apparently, we are mainly at the stage of Artificial Narrow Intelligence. Even so, human beings can largely benefit from it, and not just in the limited fields of high tech, telecom and financial services. Education, for example, is a sector where an AI-enabled approach can facilitate greater learning experiences through personalized learning for each student, replacing standardized classroom teaching. As such, we may be able to train our young people how to make best use of AI, rather than trying to compete with it.

With personal, academic, and professional data, such as mouse movements, eye movements tracking, and monitoring expressions to see if students are engaged, confused, or bored, collected through sensors and various Internet of Things devices, teachers can better understand individual student’s learning difficulties and learning preferences, and choose the most effective methods to motivate students with the assistance of deep learning algorithms and prescriptive analytics.

McKinsey Global Institute, a business and economic arm of consultancy firm McKinsey, pointed out in a recent report that high tech, telecom and financial services are leading the adoption of AI. Outside these sectors, however, adoption remains low. It is always hard to change the “status quo” mode of working which appears to have been progressing smoothly for some time. Therefore, amid all the technical fantasy afforded by AI, human factors still remain the last frontier to be overcome. As a start, the mindset of government officials and educators has to change now.

Demis Hassabis, the mastermind behind AlphaGo which defeated the world Go champion earlier this year once said, “If we can deploy these (AI-enabled) tools broadly and fairly, fostering an environment in which everyone can participate in and benefit from them, (then) we have the opportunity to enrich and advance humanity as a whole.”

Issues such as these will be at the heart of debates at ITU Telecom World 2017 in Busan this September, including AI in smart cities: power, potential, ethics and education

 

What is the challenge?

As we are enjoying a fantastic summer, we are all frustrated when we plan to meet friends in the city centre and end up in a traffic jam. When we finally reach the centre, we look for a terrace to benefit from another sunny day, but the experience is not optimal: the constant flow of cars next to us makes the environment very noisy, and we worry about air pollution levels. So the challenge is the following:

How can we improve the accessibility and attractiveness of our city centres?

The city of Fribourg, Swisscom, and the engineering company Transitec decided to address this challenge using sensor and mobile data.

We have focused on a specific and important district of a city, the train station area. Generally in the centre, as a tourist, it is the place where you gain your first impression of the city; and as a citizen, it is the place you pass by most often.

In the past, the train station area in Fribourg looked like this:

A couple of decades later, this is the way it looks now, allocating a lot of space to cars:

And here is a possible design to make it more attractive:

How do we transform?

Even if it seems obvious that the changes will make the city more attractive, merchants are concerned that accessibility will be reduced as the flow of cars will be limited and parking spaces removed. It is a fair concern, and that’s where an informed decision-making process, leveraging all available data, can help.

Can we make the centre more attractive and at the same time maintain or even improve accessibility?

What measurements do we need?

City centre accessibility is affected by three types of traffic: exchange traffic, internal traffic and transit traffic. The last one, people going through a city without stopping, is the critical one in that it is unnecessary for the city. This traffic should be steered away from the city centre.

Therefore, we need a cost-efficient way to measure transit traffic. More exactly, we need to have an indicator which focuses on transit traffic by car, since transit by public transport is fine.

To find this insight over a long period of time and with a high level of representation, we need to combine three sources of data.

Which source of data should we use?

We can combine three available data streams:

1. Road sensors, providing an accurate measurement of the volume of traffic and differentiating between cars and buses.
2. Bus sensors, counting the number of passengers on a bus.
3. Mobile phone traces, providing information on the origin and destination of passengers and drivers.

All interactions between the phone and the mobile networks are captured. The anonymized and aggregated data are transformed into the traffic indicators we are looking for: exchange traffic, internal traffic and transit traffic. Our algorithms classify the traces and provide the correct proportions. We then calibrate it with the road sensors and adjust it with the public transport data to calculate the indicator we are looking for.

The visualization of this flow demonstrates the challenge that cities face during peak hours, as you can see in this video demonstrating the movement of people in the city of Zurich from 1.00 a.m. to 8:30 a.m.

What is the outcome?

By combining and analyzing these three sources of data, we obtained the following results for our project: there are 120 000 cars entering the city every day. For the period considered, 48% of this traffic consists of cars which go through the city without stopping more than 30 minutes i.e. transit traffic.

This new insight is enabling the city to convince the population to execute this urban transformation whilst taking actions to reduce transit traffic. The action to reduce transit traffic includes the construction of a new road to better connect the south of the city to the highway entry in the north.  Furthermore, traffic light management optimization can also support steering this traffic away from the city centre.

Developing a hyperawareness capability 

By leveraging the data generated by our connected environment, cities can:

• Understand better the traffic dynamic
• Communicate their learning to citizens and increase project acceptance
• Measure the impact of projects
• Adjust their project along the way based on continuous monitoring

Cities that are embracing this new hyperawareness capability are well positioned to reach the Sustainable Development Goal defined by the United Nations and improve the quality of life of their citizens.

Finally, looking to the future, additional data can be added as our environment becomes more and more connected:

• Our infrastructure is generating more data – for example, Swisscom and partners are currently deploying 300 CO2 sensors (https://carbosense.wikidot.com/) in Switzerland
• Vehicles are becoming increasingly connected and full of sensors
• In addition to mobile phones, people are starting to carry wearables that provide additional data about their health condition and environment, including the impact of pollution, security, noise, and stress levels.

A dedicated session – Data flows, policy and security: the role of data in smart digital transformation –  at ITU Telecom World in Busan, Republic of Korea, on 26 September, will address some of the hurdles and accelerators when executing a data-driven digital transformation such as the one in Fribourg, comparing insights, case studies and experiences from around the world.