Amazon: business, or machine?

In the end my interviewee got cold feet, not wanting to risk association with commentary about her new company - she had just joined Amazon at the time of interview.


Roughly week three, she’s currently a manager in what’s called the ‘middle mile network’ (“every mile has its challenge”), so she participates in all projects - “changing routes, adding new routes, launching new sort centres.”


“Packages go from warehouses to various distribution centres – there’s a very large network all over Europe. Holidays are quite the challenge when it comes to predicting volumes for each node! Other challenges are avoiding bottle necks, and contingency planning.”


She has a machine learning algorithm background, so even though applied ML is not her focus at Amazon, her mind is still tuned in with the various elements of logistics that could be improved by it.


Drawing on prior experience, she gave me a few examples of what machine learning can be used for in the industry.


“For example, forecasting volumes is a huge part of logistics: we try to predict which sort centres will be off or on, often based on political or natural disasters.”


“Volumes needs to be re-routed correctly and at low cost, as fast as possible. It’s a very broad challenge. To apply machine learning you need to gather data to understand variations and predict what’s likely to happen next. You incorporate everything, all data sources!”


“We would make forecasting as accurate as possible, connecting with various government websites – where there’s mismatch ML improves the process – and even anticipate natural disasters. We use RPA (robotic process automation) wherever possible to remove manual processes.“


When she mentioned natural disasters, my mind immediately jumped to climate change; I would love to be a fly on the wall of an Amazon climate prediction meeting, where black and white decisions still need to be made; sea levels might rise and might immigration could happen, but consumers still need their plastic delivered.


Delivery is about cost, promise and speed. Automation helps all three.


“There are a lot of robotics at Amazon, it’s a part of it! There are robotics in some warehouses.”


I recently saw an ad (a design mock up) for an Amazon hot air balloon that glided through the sky with a swarm of drones coming and going through holes with the hive mind coordination of bees. It doesn’t exist yet (yes, I asked her anyway), but we are also in an age of rapid robotics advancement. This stuff is not that far away.


“The smallest margins - at that scale - translate to colossal differences.”


I’ve come to see amazon as a business of margins, it doesn’t matter what output – find anything that you can slap a margin on and then employ scale and market dominance to grow those margins. This appears to be their model, anyway, and it strikes me as having the same ruthless effectiveness of a virus.


Speaking again of her experience at her previous firm - also in logistics: “one of the major problems is the data because, honestly, for algorithms to do their job, the data has to be absolutely accurate. When it comes to the virtual flow of data in our systems that’s not always the case. One thing doesn’t get scanned - you get the idea: virtual flow affects physical flow.”


She qualifies that she hasn’t seen any such missing data at Amazon, which is incredible if you think about it.


There’s a detective element to the human’s role in logistics management: “that doesn’t look right” is a person’s value.


“You need to know the algorithms to know when they’re suggesting something really crazy! At Medtronic, around holidays, predictions were not always that accurate!”


I ask if ‘utopia in delivery’ is achievable?


“Oh, I think so yes, we’re developing fully monitored and automated processes to optimize every aspect of delivery.


To answer my question with more detail, she started talking about ‘business rules’ and I confess it took me a while to understand exactly what she was talking about! As a data scientist, she deals in ‘rules’ so that’s also how she thinks of business itself. For me it’s more nuanced, a bit colourful, but inputs, programs and outputs does seem to be the way business will be, erm, processed in the future (exclusively).



“Algorithms won’t replace business rules until you sort of get the AI to recognise itself.


Algorithms would create the business rules with business use cases, on top of which you place process rules.”


Algorithms on algorithms on algorithms. From her perspective, the trajectory is to put AI at the core of the business decision making itself.

“Currently the logic [of a business] is written by a person. That layer of AI doesn’t exist yet. It’s the business rules that generate the algorithms. From the data, then there is the next step of creating features.”


So her view is that Amazon will strive to automate even the isolation and addressing of business use cases, for machine designed business rules to then create process rules. In other words, a fully automated company.


With such powerful corporate entities, where will government end and private infrastructure begin?


I ask if this is basically a trajectory to the reduced use of a human being in general, silently wondering if she’s included herself in this category.

“Yes! But you could put them in other tasks, more creative tasks.”


I suppose her job is highly creative, in a technical way, but I’m not so confident creativity will always remain the remit of people. Entertainment or art as our future; the movie Wall-E gets more accurate by the day. She gives high level exec jobs as examples of the last to go, but I’m not swelling with confidence.


I ask whether people at amazon are scared of losing their job to robots (I’m not even being subtle here with my implication), but her confidence is unwavering: “I don’t think so” she puts simply, “they see automation as something that adds to their productivity.”


I’m walking on eggshells here people. What’s the trend, I ask, more humans at Amazon proportionate to machines, or less?


“The increased productivity of the company will result in an increased ability to afford many more people, for example all trucks are driven by people.”


I ask if she’s seen those Tesla trucks, they look awesome (and highly capable). It would take just one legislation - self driving cars - to replace and entire fleet of drivers practically overnight.


“Well, there are a lot of ethical questions”, ah we’ve arrived at the crux, the Trolley problem comes up, “you have to have a person, I think; would you be ok if you knew your loved one was killed by a machine? I really want to see these real world results, not just these test cases.


It’s no doubt a massive ethical debate, where do you come down on it?


The argument for statistically less accidents is so obvious as to be blinding; do we really want the fates of, even if it’s fewer people, human beings decided by an algorithm? Most people intuitively feel that the level of your involvement fundamentally changes a moral implication.


Will we ever be happy being detached from the decisions of who lives and who dies - will we even have a choice?

30 views0 comments

Recent Posts

See All

Machine learning will leave no stone unturned

  • LinkedIn

View Privacy Policy  © Machine Commons 2018