Apparently, AI (artificial intelligence) increases value from 62% to 128% for the travel sector and to 89% for transport and logistics*1.
Let’s start with some basics about what AI is. Subsequently, I will take a helicopter view and elaborate what startups can offer regarding infrastructure and the savings to be made for the pocket AND environment.
It is a culture shock stepping into a startup for the first time, I must admit. Nonetheless, wherever I see profitable and sustainable opportunities, I am determined to pioneer and follow my gut feelings.
Amazing to observe smart ‘Einsteins’ working in Tech4Good startups and Hubs. No slow-moving paper files on desks or bureaucratic meetings.
Instead, data scientists with a high degree of sustainability consciousness are addressing business and societal challenges. They collect, calibrate, model, test and validate artificial intelligence (AI) using Machine Learning, based on public and company datasets.
AI is like a sweet honey pot, still to be discovered. It can automatically identify time, spatial, energy or CO2 related opportunities. It can measure and predict environmental degradation of vital assets.
But the honey tasted bitter-sweet… How on earth can traditional stakeholders trust AI located somewhere in the cloud? They should first understand AI before any investment decision can be made.
Understanding a bee
Transportation will increasingly be impacted by AI technologies such as Bee Colony Optimization (BCO), Artificial Neural Networks (ANN), Ant Colony Optimiser (ACO), Fuzzy Logic Model (FLM), Genetic algorithms (GA), Simulated Annealing (SA) and Artificial Immune system (AIS).
Stop, hold on! This expert mumbo jumbo talk may be true, but it is too complicated.
It is flabbergasting to observe that startup ‘language’ is not always understood by their peers in established corporations and authorities. Even if they address the same infrastructure challenges.
It goes also the other way around. AI and data scientist should understand (traditional) business processes. Otherwise, no sweet ‘honey’ can be harvested on both sides…
Daily layman language which may help…
What is all the fuss about AI and robotics? Though robotics and AI are two different technologies, robotics does lend itself well to explain how AI works. And although robotics doesn’t relate directly to infrastructure, they do address the same socio-economic challenges.
AI is a computing technique that can train robots or machines to mimic human intelligence. It can recognize objects and assets, temperatures and humidity, sounds and languages, images and text, even emotions. It can plan, learn and solve problems. It can give meaning to data (insights) for decision makers.
Let’s use a simple, recognizable, daily situation.
Imagine yourself ordering a coffee in a restaurant. A sexy waitress-robot approaches you slowly, sensually…
But you don’t know how to engage with her to order a coffee. You have no clue what types of AI-enabled robots exist. You are only familiar with AI because you have seen movies such as The Terminator, The Matrix and I, Robot.
Type of AI robotics
Reactive robot – uses rudimentary AI only to react to a limited number of standard orders. Based on the way we have programmed or instructed a robot (e.g. a washing machine). This AI can’t use past experiences to make decisions. Ideal to perform functional tasks based on a ‘fixed coffee menu’ only.
Limited robot – uses AI to make decisions by recognizing patters in data. Imagine the coffee waitress robot viewing and processing a tremendous number of pictures of optional coffee menu examples as guidance to help customers. Any new situation e.g. a new, trendy coffee order or an unexpected event in the restaurant will be dealt with by computer programs attached to the robot’s memory.
Theory of mind robot – uses AI (or AEI – artificial emotional intelligence) to comprehend and react to emotions. Robots can accurately recognizes 7 emotions: surprise, happiness, sadness, anger, anxiety, disgust and neutral state. Think of a client behaving suspiciously in coffee shop, robot can immediately warn security services.
Self-awareness robot – or reflective robot – can recognize and repair itself. It uses self-awareness (!) to adapt to different situations and tasks. Imagine a new ‘born’ AI robot observing its environment while figuring out if it is supposed to be a multi-task worker, a vehicle repairman or a self-repairing coffee robot. Based on Deep Learning algorithms. Tested by Columbia Engineering (Jan. 2019).
Be informed researchers may use different definitions to explain AI robotics more in-depth. All good and probably true. Transmedia prefers to use simple layman language to bridge the AI gap towards potential users.
AI and algorithms: explained to a child
Algorithms are symbolically spoken fast growing, intelligent children who learn how to distinguish ‘good’ from ‘bad’ (data).
Deep Learning, a subset of Machine Learning, is like a parent teaching children statistical and research technique how to independently build analytical models (without being explicitly being told or ‘programmed’) around data to solve problems.
Ground based ITS tools are a nice starting point, but deep learning techniques can advance it by processing large volumes of data e.g. captured by satellites, airplanes or drones to offer new insights.
Traditional techniques can’t crush all those data to get relevant insights. In contrary, AI ‘likes’ uncertainties. It models a connection between cause and effect in various real-life scenarios.
AI is a means, not an objective
Corporates and authorities should not embark into AI just because it is trendy. Potential users should have prior in mind specific situations in which AI could automate processes. Important lesson: AI does not give solutions, it helps you to find solutions.
Your company or sector should be digitalized and data rich. Why? It takes thousands of examples or training data to learn of all the characteristics related to a city, or transport movements. Also, there are many lacks, gaps, noise and uncertainties within ‘dirty data’.
Today, 80% of data seems to be unstructured. There are methods and filters to clean data. But: all this costs time at the beginning.
Your problem fits in 5 situations
Most infrastructure related decision makers wish to be informed or alarmed in 5 scenarios or situations. Obviously, there is a great variety of cities, industrial sectors and supply chains with different requirements.
Generally, stakeholders share a common set of challenges:
- Did, is or will a predefined change take place in a city/environment/infrastructure, or not?
- Did, is or will something strange or suspicious take place, not according to normal settings?
- How many changes or trends take place impacting city/environment, or finances, sales?
- How are groups organized in city/crowds/traffic?
- How can I maximize performance by automatically learn and improve a situation within a city/environment/infrastructure?
Let’s take Q1 as example while obtaining quick, basic insights. Data input for Q1 could be speed, flow and occupancy. Auxiliary data (bases) could deliver additional data about location, events or weather conditions. A result could be congested vs non-congested (= classification algorithm).
Monk’s job being automated
Recognizing small objects like buildings or vehicles in large quantity of data images can be executed by satellite imagery analytics. A single satellite image could cover over 60 km2 and millions of pixels. Hence, AI can turn a monk’s job in an automated data scientists’ job to train, evaluate and test millions of images.
Quick case: measuring train station on the move
Data from Sentinel-1 satellite illustrate that parts of Oslo train station are sinking by 10–15 mm a year in the ‘line of sight’ – the direction that the satellite is ‘looking’ at the building. This translates into a vertical subsidence of 12–18 mm a year. Apparently, the new opera house (white structure) has not moved. Perhaps Pavarotti was inside, who knows. But what do you know about your port or railway infrastructure?
Quick case: assessing corridors
Satellites can assess the state of different fixed transportation assets in corridor perspective. They monitor how it evolved over time. Also, land-use information and population distribution around key access points can be provided.
Quick case: monitoring coastal water pollution
Near real-time monitoring of water became feasible since recently. E.g. automatic warning of sudden changes in water supplies, disappearing of water due to sudden hot periods, early warning for looming flooding.
Quick case: solar potential
Cities are eager to measure CO2 emissions or heat values per area and compare it over time or between districts. Also building owners want to know how to save energy costs and lower tons of emitted CO2. AI can even measure Air quality (emissions).
Green savings through insights
Transport and mobility players are searching for smart and green growth opportunities. Operators want to offer more than just being 1% cheaper, 2% faster or claiming to use 3% less fuel use than the competition.
On average, AI increases value from 62% to 128% for the travel sector and to 89% for transport and logistics *1.
Also, industrial federations and membership platforms search new instruments to create significant impact in favor of their constituency and the environment (‘Paris’ climate agreement).
Startups in the network of Transmedia.online are dealing with various demands and questions of customers to cut costs while serving sustainability objectives:
“Warn us in case of anomalies near nature areas, near our assets, or elsewhere”
“Detect heat loss in our city/port/logistic centers as input for our insulation strategy”
“Calculate the exact length of our waterways, railways and highways 99% accurate”
“Quantify accessibility and friction points per transport mode to our facilities nationwide”
“Propose optimal locations for future electric vehicle recharging points”
“Make inventory of air pollutants in our city and port areas”
“Localize, monitor and plan solar panel and green roof business opportunities”
“Detect and measure the condition of bridges nationwide”
Governmental agencies, also global corporations are searching ways to cut down on inspection costs of vast areas, or great number of assets. Instead of expensive manual ad hoc inspections during years, AI could give a global view quicker, more precise and more cost-effective at the end.
Sustainable business governed by AI?
The last two decades, millions of R&D resources have been spent to obtain insights on how to use infrastructure in a sustainable way.
37% of climate and environment-related technologies worldwide are originating from the EU. All accurately executed ‘green’ projects by the book, but with a success rate of 11.6 % *2. Smarter and cost-effective instruments delivering relevant insights are required.
The climate needs smarter and stronger ‘breakthrough’ instruments and insights to change course towards the ‘Paris’ agreement. AI has started just to do that
*1: McKinsey Global Institute: San Francisco, CA, USA, 2018
*2: EU 2017 mid-term evaluation report Horizon2020