Self driving continuous delivery?!

by August 19, 2019

Your journey to continuous delivery is a lot more closely related to autonomous driving then you may think. in fact, just like continuous delivery, self driving cars are all about levels of automation. And intelligence you may ask? Intelligence is a level of automation in and of itself!

Self driving cars are all the rage these days, but they are not really here yet. If you dig a little deeper into it you”ll quickly find out that the NHTSA (yes that exists, its the National Highway Traffic Safety Administration board in the USA) has adopted what they call the 5 levels of automation ( 6 levels actually if you consider level zero “no automation” as a level), anyway – you can read about it on the NHTSA website here.

The 5 Levels of Automation in Cars.

The interesting thing to notice about the NHTSA stairway is that they actually don’t really mention artificial intelligence or deep learning at all, instead they make their distinctions based on the “level of automation” that a vehicle is capable of. So for example, level 2 is defined as “partial automation” where “the vehicle has combined automated functions … but the driver must remain engaged with the driving task and monitor the environment at all times”, at level 3 the driver is still responsible for the driving and must remain alert all thought the vehicle practically handles all the driving (Some modern day cars like the Tesla and other are considered to be almost level 3), At level 4 the car can drive autonomously under certain conditions (e.g., highways, specific weather conditions etc..) and at level 5 the driver is completely optional, or indeed, not even given an option to control the vehicle. 

Where is the intelligence?
When you begin to further explore these increasing levels of automation and think more closely of the technology that would be required to achieve each new level of automation on the NHTSA list, the necessity of artificial intelligences or deep learning technologies starts to become clearer. For example, In order for a car to be able to drive safely in a busy city and avoid hitting that crazy dude who suddenly jumped on the road,  Or at least give it it’s “best shot” better than a human driver could, the car will need to be able to “understand” the situation in a split second even if it had never been in a similar situation before, Just like most of us. This means it will need to “learn” about many things before it ever encounters these things in “real life”, it will need to be “taught how to react”… intelligent stuff. Or consider the vehicle’s ability to adjust to rain and snow etc… without delving too deep into the philosophical range I hope we all see there is hardly even a thin line between what is “intelligence” to what is “automation” in this context. The key to setting goals for us is not in an IQ test but in specific capabilities and scenarios (i.e., use-cases) that our cars can handle for us. The same logic applies, and should be considered for digital transformation and continuous delivery as a key component of it.

Humans don’t ship in containers

Surely this whole story is completely irrelevant for continuous delivery or release automation! Or is it?? My argument here is that any transformation of human activity to an automated one always follows a similar, staged, evolution. Recognizing the main levels of your particular circumstance is key to ensuring your goals are set monitored correctly in a structured way forward  that stays in line of sight. We may be a lot more dismissive of the importance and attention a continuous delivery transformation is then a self driving car because we do not have to, literary, sit in it. However, digital transformation and continuous delivery both share a progress continuum that revolves around levels of automation, entirely not dissimilar to those set by the NHTSA for cars.

The 5 levels of automation in Continuous Delivery – Welcome intelligent pipelines

Much like a level 2-3 autonomous vehicle many organizations are already boasting automation in their release pipelines and even “continuous delivery” but much like autonomous vehicles, not even  one organization in the world exists today where even human supervision of the CD process is truly redundant, or shall we call it level 5 CD? We may have autonomous deployment processes and some autonomous rollback implemented in some (high end cars anyone?) instances but when you consider true continuous delivery – from ideas to user feedback – we can’t ignore the gaps that still need to be overcome before we can claim “level 5” in human supervision over the process and particularly in quality assurance as an integrated part of “self driving continuous delivery” vehicles.  

Read more about intelligent pipelines here.
https://www.ca.com/us/products/continuous-delivery.html