Saturday, July 1, 2017

Don’t fear the AI - embrace it!

In 2017, Artificial Intelligence is here in our world. Tools such as Google's DeepMind and IBM’s Watson starting to gain real world application. We are starting to use AI to automate many tasks and alter the way we work.

There has been much debate in recent years around what Artificial Intelligence is likely to bring. Some people fear that AI will replace human workers and there will be mass unemployment. Other people fear that the machines will become so intelligent that they will take over the world.

Let me point out that all of this is just speculation, albeit some of it is an educated guess but much of it is science fiction. There is absolutely no evidence to many of the claims. However, unfortunately, some of those claims are made by some of the smartest people on the planet. This makes us all take notice. Another unfortunate scenario is that those people are now asking government to respond to this as a threat, in a similar fashion as there are calls to respond to climate change. The governments are asked to spend our taxpayers money to combat some speculative future that was dreamt up by some smart people. Now this is a cause for alarm. Not everyone has subscribed to this new mass hysteria though, there are voices of reason out there that include many scientists. This alarm inspired me to spend my time and write this blog in order to add perhaps a small extra drop of logic to those voices of reason.

I am not a scientist researching AI, not a politician or a super wealthy entrepreneur. I have been trained in computer science and did work as software developer for many years so I do understand technology. I also created a number of businesses and helped hundreds of businesses implement technology so this gives me an idea of how jobs are created and how business operates.

But the main point I want to convey is that I am not creating any new ideas here. I am simply applying some logic and common sense to the information that is in public domain.

Below are some of the myths that particularly cause me to be alarmed. I will offer some thoughts by way of “dispelling the myths"

Myth #1:

There will be mass unemployment and hardship. 

Artificial Intelligence will make many jobs redundant. I have no doubt about that. However, I believe they will be replaced with better jobs.

There are some big mathematical and logical issues with the gloom predictions which most predictors ignore. First one is that people are in control of implementing the AI. They are the ones paying for it, either as consumers, as decision makers who write cheques in larger organisations or as business owners.

So let me ask you a question: What is it that you like in your own job? Would you buy technology to replace a job you like? I personally don’t know anyone working on technology that would replace a job they (or someone else) enjoys doing. And if someone did - that business would be unsustainable. Most efforts are going into assisting with the mundane, boring or physically difficult parts of jobs. This logic is also supported by results of 2015 study by Deloitte economists into the relationship between jobs and raise of technology since 1871. The study has shown a clear increase in the number of jobs created as the result of the rise of technology and innovation. You can find the findings by clicking here.
However, there are smart people like Peter Hinssen who tell us to not extrapolate what we know today to try and explain tomorrow. What Peter really means is not to extrapolate data trends into the future.

Peter has shown us graphs where Telecommunications providers were looking at the up take of SMS in 2010.

And if you looked at that graph and extrapolated the future you would see exponential growth in SMS usage. However new social media based messaging systems arrived and spoiled the predictions.

Peter’s example serves a great warning to anyone who uses pure statistical data to arrive at conclusions. I certainly never agreed with statistics as explanation for everything and fully agree with Peter on this count.
However, this doesn’t mean that we should also throw away a-priori logic and everything we know. After all, Peter himself uses logic and what he knows today to paint a picture of the future.

So the logic that I propose is this: While there are plenty of professions that used to exist and no longer do,  to date we have seen no evidence of technology causing mass unemployment. Fact is that to date there has been more jobs created as the result of technological progress than we ever had before.

The mechanisms by which humans implement new technology only support implementation and ratification of new technology as new jobs are created. The more people have money to spend the more new products and services are created which in turn fuels the research to make them faster and smarter. Those same mechanisms in play will also slow down technological advancement when old jobs are automated.

If people were becoming unemployed “En masse",  the first thing that will happen is economic downturn. That causes companies to stop, or substantially slow down investing in technology. I witnessed this with the GFC. As soon as unemployment increased by a few percentage points, chequebooks closed and companies stopped investing in new technology. This may sound counter-intuitive, but this did happen, and if you think about it for a bit, it makes sense. When people are unable to earn money, they turn to producing the basic necessities by themselves. Same with companies - when there is a business downturn, they try and get their staff to do more in-house than outsource or invest in capital purchases.  In other terms, if people don’t have money to pay for the AI delivered products and services, then they will not use those products or services and perform the work manually for themselves instead. This will in turn cause companies implementing AI to go out of business or stop investing in AI.

This means that when workers loose their jobs the economy and rate of technological progress slows down. It will slow down enough to let those workers re-tool and get back into employment.

This means that "mass unemployment and hardship” with “people unable to earn money” is impossible as a result of technological progress. The a-priori logical conclusion here is that the technology advancement goes hand-in-hand with increased or full employment.

So in that case what likely impact will technological progress make on jobs?

Let’s consider the car industry, and dispel another myth while we are at it.

Myth #2:

Autonomous Vehicles will make anyone who is earning money driving unemployed and unemployable.

When talking about mass unemployment the doomsday advocates often refer to the taxi and logistics industry. They say thousands of taxi drivers and truck drivers will loose their jobs overnight and will be unable to recover and retrain for years. Those predictions could not be further from the truth or reality. Here is one opinion that supports me in saying "Driverless Cars Will Create New Jobs, Not Destroy Them". Let me explore why in my own words.

Before we look at the future, let’s have a quick look at the recent past and the major disruption in the taxi industry caused by Uber. Uber managed to lower the costs of taxis by implementing technology that allows drivers to deliver better service to their customers at a lower price. This caused some taxi companies to loose market share. What did the drivers do? Most of them comfortably transitioned from being employees to becoming business owners. They bought a car and use that as a cab and use Uber technology to provide marketing, sales and billing. Very clever.

We all know that no driverless car has been approved or put into the general population consumption anywhere in the world as autonomous vehicle. That is because approving that kind of technology require cutting through a lot of government red tape and takes time. Let’s say a manufacturer (for example, Tesla) creates a fully autonomous vehicle that is starting to go through government approval to allow it to go on the road. What happens next? One type of vehicle gets approved. Let’s say it is a light passenger carrying vehicle. If a taxi company will then decide to invest into such vehicle, will they change from cars with drivers to driverless overnight? Highly unlikely! What they are likely to do is buy a small test run of vehicles to try. This trial likely to last six-twlve months. Uber is also unlikely to invest in a global fleet of vehicles overnight. People more likely to jump in ‘boots and all” are the uber drivers. All they need to do is buy one vehicle and they are sorted. Why would Uber even contemplate investing in a fleet of driverless vehicles, when their traditionally disruptive model relies on the drivers to invest in global fleet available to Uber? By letting drivers invest, Uber can become a global force dominating driverless vehicle industry almost overnight. And what will happen to the jobs? There will not be “thousands unemployed” as the drivers who used to be behind the wheel now will enjoy the cars doing the job for them. All they need to do is administer the business.  Will they need to retrain? Eventually yes, but not immediately. They already made the transition from employees to “independent business owners” - so now the journey will begin to learn not only how to manage one vehicle, but how to grow their business and add more to their fleet. And nobody is in a better position to do that then those drivers as they already know their area, know their customers, understand the industry very well.

As the prices will start sharply falling for the taxi rides a couple things will happen:

People will stop buying cars. 

Why would you own one if you can get a driverless uber instantly at your door taking you to your destination for less money than owning a car yourself?
This in turn will increase demand and let more people put more driverless cars on the road. This in turn will increase the employment in the sector as those drivers will start hiring administrators, marketers, sales people, software developers...

People will start using cars for new purposes

Why go to the supermarket, when the delivery of your shopping costs you the same as the cost of driving yourself to the supermarket and back? We will see more things couriered where they were not before.

All of that will create more jobs in the taxi industry. Sure enough, the “drivers” will need to re-train. But like many start-up owners, they can mostly re-train “on the job”

A similar model of disruption and change can be built for any other industry.

The industry will change with AI, the way the job is performed will change and perhaps even what we call each role will change too. While the mundane and boring parts of the professions will be automated, the interesting and human parts will not. Also newer technology is more user friendly, easier to use and requires less training. This means that as the jobs change, they will often favour a more mature workforce with industry knowledge and soft skills.

Myth #3:

Machines will eventually be able to do every single job, humans will be fully “redundant” and people will be unable to earn money.

Before examining wether or not people will be able to earn money, let’s examine what “earning money” really means.

We, humans, trade fruits of our (human!) labour with each other by exchanging goods and services. When the other doesn't have the exact product or service we need right now, we use money as the store of value for human labour. So money represents human labour.

We use tools to make our jobs easier, however the tools are in themselves a store of value of the labour that went into their invention and production. Eventually that depreciates and the impact of tools on value we produce is a race to zero.

No matter how smart the machines are at doing our jobs, they are at the end of the day mere assets of humans. The cost for products and services “provided” by machines, in a competitive world, will be a  race to zero.

If the machines could do everything, than everything we need ultimately will be provided for free and we will have a life where we don’t have to do anything and have everything.

However, machines will not do everything. This is not because I said so, but because by definition, some of the tasks will not be able to be performed by anyone other than humans.

Machine is not capable of performing human-to-human contact because definition of such activity is based on human participation.

Right now service industry is experiencing labour shortages and in order to satisfy demand that industry will have no other choice but to implement AI of some kind - sooner or later.

I am certain that AI will fill in superbly and many, if not all, currently available service jobs will include an AI of some kind.

However, humans do biologically crave human-to-human interaction. I don’t believe we will ever be satisfied without human interaction.  This mean there will be a need for some kind of human-to-human interaction services. As such, by definition, those service will be performed by humans.

So if you look at it mathematically and extrapolate the numbers, the cost of AI / machine workers is a race to zero. Human-to-human service will always bare a cost, that is infinitely higher than the cost of AI service. However, as AI is unable to replace human in those circumstances “by definition!” then the demand for the service will always be there. The demand will eventually grow to fill the available supply. Since we will get all the AI services for free (or nearly free), we will be happy to spend "all our money" on the human services.

Also, at the flip side of it, money is by definition is a store of value of human labour. Trade is by definition a human-to-human value exchange. Now before you start telling me that there are machines trading equity and so on, please consider that all the ownership of value ultimately is stored with humans. This means those machines are acting on behalf of their human masters and are simply the “tools” of the job performed by humans.

Another, very important part that machine does not do is - it does not have initiative to innovate or explore new human frontiers.

Inovation in the humanity space and exploration of the new human frontiers is again, by default are human actions and can not, by definition, be replaced with AI.

Since we will always require human interaction, there will always be jobs for humans. If humans didn’t need to do anything else but just that, it will be enough to keep everyone employed.

Myth #4:

Machines will rule the world and will quash us if we get in their way.

All machine code is written by humans. The technology is sponsored by someone.

Will you, or anyone you know, fund a machine that has in it’s algorithms somewhere a capability to destroy human life?

Perhaps if the machine is created for military purpose, yes, maybe. However, we already created military technology that is capable of destroying all life on earth. So far we are able to keep it “in check”, but who knows what the future brings. The point is, wether it is military AI or Nuclear Bombs - if it gets into the wrong hands we are all doomed. If there is one thing I fear, it is the autonomous AI driven machines that are given the mandate to kill humans.

However, can you conceivably imagine a commercial company funding development of such machine, without a military chequebook? I can’t.


No matter how smart AI gets, there will always be something for people to do. MIT recently published a research paper that also agrees with this thesis. You can find a link to that paper here. While industry definitely needs to take notice, on a citizen level there is nothing to fear. There is certainly no need for governments to spend money. I am not saying governments should take no notice of the situation. There will be plenty of moral and ethical dilemmas to consider, including AI in military applications.  The legislation will need to be adjusted to deal with those. But that is nothing more than “business as usual” for the governments.