Having discussed the current approach to Artificial Intelligence and noted that we can create smarter system by endowing them with Machine Learning (Machine Learning is based on Artificial Intelligence but not all Artificial Intelligence embeds Machine Learning, however I suspect that this latter will become more and more part of any artificial intelligence application. Isn’t learning an integral aspect of intelligence?) let’s move on and consider a few real applications of AI, stripped from the hype that is claiming AI everywhere, not because it is there, nor because it is needed but just because it is a cool marketing tool.
There is a lot of talking (and hype) about application of artificial intelligence in the mobility area. However, in m
any cases what is being done is some form of advanced data analytics and not real AI. Remember that true AI today requires plenty of data and data crunching capabilities and in most situation it is not economically viable to have this kind of capabilities at the edges (e.g. in a terminal/a vehicle). At the same time the connectivity fabric has not reached the point where massive data communications is both reliable, cheap and ubiquitous. The advent of 5G (of mature 5G; not the one being deployed here and there) may change the landscape, let’s say by 2025. Tech evolution will also make edge computing much more viable.
In the mobility area we have the basic required ingredients: huge amount of data and continuous changes, what it is needed to extract meaning and initiate smart action.
Here a few examples of real applications, today, leveraging at least a bit of AI:
- Self-driving vehicles (including level 3 upwards) where the detection of obstacles and evaluation of threats through image recognition and understanding using computer vision is making full use of artificial intelligence and the software benefits from machine learning applied to the thousands of situations faced by different vehicles;
- Best path between a and b (or any number of points, like in the salesman problem). There may be many factors involved and leveraging on experience is often a good approach. The complexity in city traffic and the variety of constraints make the use of AI effective. Several delivery companies, like Uber, Foodora, Doodash, Grubhub, are using AI to optimise travel time and delivery of food, taking into account what people are likely to order at a particular time, with that sort of weather and what restaurants promptness might be …
- Traffic orchestration through smart signals. Waiting for the time when smart vehicles will be able to talk with one another and self-orchestrate their path and velocity to achieve the best use of the available resources (roads, parking space, recharging station…) a few companies, like RapidFlow, are already offering smart signals solution with an AI orchestrator that looks at the global picture (traffic volume, trend, weather, events, resources or lack thereof) and at the (ranked) needs of drivers/passengers/goods.