Posted by

Brian Keng

on July 12, 2018

Share to

More Posts...

AI: The Next Evolution of Automation

by Brian Keng | July 12, 2018

At Rubikloud, we believe that AI is causing a monumental shift in the way retailers run their businesses. However, you might be surprised that it’s not their fleet of flying delivery drones, nor their checkout-less stores that are the most impactful to their business but rather their uses of AI in their core business.

We think that the real impact of AI for businesses will be beneath the surface, not with talking robots or self-driving cars, but rather for the thousands of other processes that affect businesses such as the supply chain, merchandising, and marketing. The underlying value here is not the technology of AI but rather the automation it provides. Using this lens, AI is nothing more than a technology enabling the next evolution of automation.

The History of Automation

Automation has a long and rich history dating back to the first feedback control systems in water clocks but perhaps one of the first modern uses of automation was with the steam engine and the industrial revolution in the 17th century. This brought about the need for the first automated control systems to regulate critical factors such as temperature and pressure. From that point on, the need for more sophisticated control systems only grew with advances such as relay logic and the electrification of factories.

Early Computing: Goodbye to the slide rule!

Beyond control systems, the 19th century brought about the first mechanical computers with Babbage’s difference engine and even conceived perhaps the first general-purpose computer, a century ahead of its time. The 20th century brought about the exponential leap in computing power from analog to digital computing and from vacuum tubes to transistors. The different computing devices were used in pivotal moments in history, such as; Turing and his work on cracking the Enigma machine in World War 2, the development of the atomic bomb, and landing a man on the moon. Automating these tedious, error-prone and highly complex mathematical calculations were one of the earliest productivity gains for early computing. This started the long journey to more computationally powerful and, counter-intuitively, physically smaller computing devices.

One of the earlier computers, a drawing of Charles Babbage’s Difference Engine. (Wikipedia)

The Microcomputer and the “Killer App”

The 1980s brought about the “microcomputer” — the first generation of computers that were widely sold for personal and small businesses use. This also brought about the “killer apps”, early software programs that unlocked the power of these microcomputers. Apps such as word processors and spreadsheet programs, created demand for the underlying hardware because of the massive boost in productivity, automating and digitizing what was once the realm of the physical: paper and mechanical typewriters. This democratization of automation allowed everyone to have access to the same tools, not just those who could afford a legion of paper-pushing secretaries and analysts. These killer apps were just the beginning of a trend of software “eating the world” by automating mundane and routine tasks.

A collection of early microcomputers. (Wikipedia)

Connectedness and the Web

An assortment of networking related research throughout the 1960s and 1970s, led to the precursor of the internet known as ARPANET in the 1980s, eventually evolving into what we now know as the internet in the 1990s. This global network would eventually enable every computing device on the planet to be connected and with it the potential for further automation.

This next step in the automation revolution removed one the last barriers to unlocking the computing revolution: having them talk to each other. No need for those paper printouts, or fax machines, you now could just let the computers do all the work. Email and standardized protocols opened the doorway to a huge plethora of innovations from e-commerce to social media, uprooting a lot of traditional businesses still stuck in the physical world. All of a sudden, the internet democratized information. Now any piece of information you could imagine was at your fingertips (including how to write software to automate gathering and processing it)! The next logical step in this evolution in computing was not to just make them faster, or letting them talk to each other, but rather to make them “smarter”.

Rapid growth of internet users since the 1990s.

The Next Step in Automation

AI’s birthplace is frequently attributed to the Dartmouth Conference in 1956. These early AI techniques relied heavily on symbolic techniques such as logic, search, semantic graphs, and hardcoded rules. This field has had over sixty years of research, gone through two “winters” of overoptimism and extreme pessimism, and now has had a resurgence in the last few years.

The big difference this time is AI, or rather the modern incarnation of it with machine learning, is capable of building truly smart automation systems. Whereas in the past these machines could only learn to solve toy problems from data, this generation of AI has been supercharged by the growth in raw computation power and connectedness of our modern computing systems allowing them to “learn” complex tasks from data. Many of which are tasks that are beyond the capabilities of humans to explicitly build. This overcomes a huge hurdle of past iterations of AI techniques by removing the last manual bottleneck (humans) in automation, and simultaneously introduces new challenges ensuring that the machine “learns” the correct thing.

These new learning machines necessitate a new way to think about building and applying them compared to traditional software systems. We can no longer rely on a human to build all the intuitive “human” aspects into the system, instead we need to carefully curate the data and build harnesses that encode all the realities not present in the data. This is even more critical when we’re relying on these learning machines to automate tasks that can affect business critical functions. Machines may be doing the learning, but the humans have even more work to do in building these systems.

This brave new world of AI automation is still in its infancy, and like the advances in automation before it, AI will have a massive impact on every industry. Rubikloud is working with enterprise retailers around the world to navigate this next evolution of automation by building intelligent decision automation products that help empower retailers to leverage this new technology in their business.

If you get a kick out of teaching machines to learn and enjoy automating everything from your unit test to your toaster, Rubikloud is always looking for ambitious and curious data scientists and engineers: