Big leaps in A.I are happening so fast that they are hard to keep track of. You may have heard of an A.I milestone which occurred in June when IBM’s Project Debater drew with human debaters. Or that a program built by two Carnegie Mellon researchers beat several top poker players over a 20 day contest. But why are these big events? And does this mean that A.I could really be trusted to make decisions to boost revenue in mobility solutions for the likes of car-sharing and on-demand micro-transit?
One of the critical techniques used by the poker A.I known as “Libratus”, or ‘Balance’ in Latin, is reinforcement learning. Basically the researchers let a deep neural network (similar to those used in self driving cars) try out different strategies randomly and then reinforced its good plays. Over millions of plays this a large network can learn to solve almost any complex revenue optimisation problem. Even when there is significant uncertainty, just like there is in car-sharing operations.
Does this mean that computers will take over high level business decisions from humans in our rapidly changing industry? Of course not. But it does mean that we can now trust advanced A.I techniques with complex resource management tasks that are increasingly found in on-demand micro-transit.
We’ve had significant success here at Good Travel Software using the latest A.I techniques, including reinforcement learning in the management of vehicle relocations in large scale fleets, dynamic pricing in floating car sharing schemes and in pick up route optimization for dynamic buses. But reinforcement learning is a special technique that can be applied to almost any problem that comes under the general heading of logistics. The solutions provided by our drop.car system marks a huge leap forward in car-sharing technology.
What technologies make an AI implementation feasible in the mobility industry?
You might think that A.I is a field left only to the Googles and Facebooks of this world. But there has been a significant effort by these big players to make their technologies open to researchers and even easy to implement on a large scale. Some of the key technologies that made it feasible to integrate a robust scaling A.I solution into our business platform were:
- The open sourcing of Google’s highly scalable A.I library Tensorflow
- The rise of micro services and containers such as Docker along with container management systems such as Kubernetes
- Elon Musk’s OpenAI platform which regularly publishes new research, algorithms and simulation frameworks
We’ll leave an explanation of the significance of each of these technologies and platforms for a later blog post. But the point is that world class A.I algorithms and technologies are mostly open sourced at the moment which is extremely exciting for both researchers and companies looking to use these latest technologies. Big innovators such as Uber are also using Tensorflow and are just as excited as us about reinforcement learning.
What does this all mean for the future of the transport industry?
It means smarter A.I systems will soon run our logistics, pricing and transportation networks using robust strategies that take into account the uncertainty of day-to-day business operations. This will improve fleet-utilisation, user satisfaction and in the not-so-distant future these techniques will become the cornerstone of autonomous fleet optimisation. This will mean huge growth down the line in the car-sharing industry.
Our fleet-balancing software drop.car is just one of our many flexible yet powerful car-sharing soltutions designed to improve your business.
To get a glimpse of the future contact us to arrange a demo.