So I’ve been getting a lot of “Hey Congrats on that thing you won, um so what was it???”. This speaks to how much confusion there is around Machine Learning and Artificial Intelligence, and I have shared in this confusion. So allow me to simplify what just happened.
I attended the AWS Summit last week, it’s a free trade show that provides access to learning sessions around Amazon’s cloud offerings. These are the building blocks that are changing, improving, and expediting the way we build technology. One of the sessions was an introduction to #machinelearning using a toy car called #deepracer, a reference to “Deep Learning”.
In the two hour session, the speaker introduced us to a model called “reinforcement learning”, and utilized slides depicting a good dog getting a reward, and a bad dog not getting a reward. In machine learning, this reward system is implemented using points. You get more points for the desired behavior, and less points for the undesirable behavior.
OK, so how does this tie into the toy car and the race? #AWS pulled a slick move and basically #gamified this complex subject by connecting the technology to something we all love – toy cars and racing. By doing this they made Machine Learning approachable by a wider audience.
To do this, AWS built a feature into their console that allows you to control the driving characteristics of the car, and build this into a Machine Learning “model”. This model then gets trained by a farm of processors in the Amazon cloud, and you can watch your car “learn” how to drive in a virtual environment.
In order to work in this environment, you do need to know your way-ish around the AWS console, and know a little bit how to code in some language. Fortunately with 23 years in tech I had both of these skills. What also helped me out was my previous experience as an automotive tech, and my love and understanding of driving sports cars. All of this influenced my approach to tweaking the parameters provided in the model.
We only had 45 minutes to play with this environment in the workshop, and after playing with the default model, and then building my own based on some assumptions, I has to sit back and watch my model train and learn how to drive by failing. I didn’t have the luxury of allowing my model to train for the full 60 minutes, so I had to cut it short. At the end of the session the presenter said that we could take our model down to the racetrack on the conference floor and race our model in the real world.
With a little help from our friends at @matillion who loaned my a USB dongle (#grateful), I stood in line at the race track and waited my turn. I had low expectations as to how my model would perform, nor did I realize there were prizes for winning, I just did it for fun. My first lap went absolutely perfect and placed as the fastest lap time of 10.43 seconds. I was shocked. Every subsequent lap didn’t do as well, so I think someone must have been watching over me that day.
With less than an hour left for competitors to challenge my time, I milled around the track and did a short interview with @Ryan Myrehn, a real world “pit reporter” AWS hired for this event. As the end of summit reception started, I was whisked off as the champion and presented with commemorative jacket and trophy on stage, and was interviewed for some blog posts and promotional videos. I didn’t have time to get nervous, it just happened so fast.
As it turns out I also won my own AWS DeepRacer 1/18 scale autonomous car, and an expenses paid trip to AWS re:Invent 2019 in Las Vegas in December where I will compete with 20+ other Summit participants from around the world. Until then, I am now sufficiently motivated to dig deeper into Machine Learning on the AWS platform.
At the start of my day I had intended on finding out more about Machine Learning (ML) and Artificial Intelligence (AI). Why? As the CEO of a software development and cloud services company, there is no denying the relevance of these recently available technologies. Although few of our current customers are asking for these features, I am certain that we will need to incorporate ML and AI into the software we develop. While it’s not my job as CEO to code or to learn these technologies deeply, it is my job to champion adoption by our team and to know we’re implementing currently accepted technology and best practices.
See you at the races kids!