Disruptive Forces: Machine Learning as Societal Disruptor (2016)
Presented by Steven Rahman to The InterAction Council, Baku Azerbaijan 2016 as part of discussions focussing on the future of work and global youth employment.
Thank you Mr. Chairman.
I must say it’s very much an honor to have the opportunity to speak here and participate in this discussion. I am very grateful.
I am not a policy maker. I am speaking today because of my experience in Silicon Valley where I have worked for large technology incumbents as well as disruptive start-ups. My focus has been on digital technology…so, please consider my comments in that context. In these remarks, I do not intend to provide opinions about other technology areas such as life sciences or biotechnology — topics which were covered so expertly this morning.
I’d like to begin with software. I have a gloomy view of what will happen to labor markets in the near term because of specific advances in technology that we have seen over the last five years.
The main reason that venture capitalists like to invest in software is that software is scalable, meaning that the unit cost of a licensed instance of software drops extremely quickly as more units are sold. For example, imagine Microsoft Word. The first unit of Microsoft Word that Microsoft sells might have a unit cost of $5 billion dollars. The unit cost of Word for the next unit sold could drop to $0.75 cents. Eventually, after enough copies of Word are sold, the marginal unit cost approaches Zero.
I am over-simplifying but you are getting the idea.
Marc Andreesen, a venture capitalist at Andreesen Horrowitz, is most famous for being a cofounder of Netscape and a pioneer of the Internet Browser. He is credited with saying that “software is eating the world.”
What does he mean by that?
First, lets look at how software is designed. It’s a combination of Scalable Solutions, Design Thinking, Agile Methodologies, and Rapid Prototyping.
Software is scalable. If you make a monster product like Microsoft Word or Gmail, a very small team might touch millions if not hundreds of millions of Users. There is a mobile game company from Finland called SuperCell. In 2014, they produced revenue exceeding 1.5 billion Euros. At the time, they had approximately 150 employees.
Techniques in software design have evolved, making software makers extremely good at getting their products into market. For example, you may have head the term design thinking. It refers to a concept in which engineers and designers design solutions with very human problems in mind. 100% of a software design thinker’s effort is focused upon solving something.
Other techniques include agile and lean methodologies and rapid prototyping.
In digital, designers manage their creation process using something called agile methodologies supported by additional technology and software provided by others within the ecosystem. Dr. Michael Porter in his work on competitive clusters has written extensively about the importance of ecosystems, so it is unsurprising that helper companies have sprung up alongside the software developers.
When developers release their software, they monitor carefully customer use of the product. User sentiment is quickly fed back into the design process. This is often called the “lean methodology” in which companies can quickly “pivot” based on changing needs of customers. It also means that developers can discover unexpected use cases of the software. And then pounce.
This methodology is in contrast to the way traditional engineering projects are completed. You will never go to the moon managing your engineering teams with agile methodologies, but with agile, you might be able to shine an advertisement on its surface.
Rapid prototyping allows teams to test their product with users quickly. They can abandon ideas and abandon features. Or emphasize unexpected uses. A quick example: A company called Burbn was founded by a small group of design thinking acolytes from Stanford in 2010. They built the application with lots of features that they imagined people would use. But they noticed something…their users focused almost exclusively on photo editing features. The features allowed users to give texture and apply filters to ordinary pictures shot with a camera phone. The team “pivoted” — meaning…they changed their focus to emphasize the photo features and renamed their company Instagram. It was sold two years after founding to Facebook for $1 billion.
This type of thinking is not restricted to the digital world.
Software has disrupted the hospitality, automotive, and livery industries. It has allowed applications such as Lyft, Uber, and Airbnb to disrupt established industries with surprising ease.
Uber has allowed consumers to remove automobiles from their personal balance sheets and Airbnb has allowed property owners to gain returns from previously unproductive assets. I think that it would be worthwhile for an economist to explore the inflationary effects caused by Airbnb and Uber as the revenue streams and savings they create are very much like a tax cut.
Applications like Uber, Foodora, Task Rabbit, and Postmates have created many temporary work opportunities for people working according to their own schedules. However, wages have stagnated around the world despite these new opportunities. Furthermore, the temporary nature of this work has created a class of people who work by themselves without benefits, without access to training, and without access to mentorship. Many of these people had other jobs — now they are driving “taxis”. There are thousands upon thousands of Uber drivers who might be in these service roles because of an absence of other opportunities.
This is an interesting period of time where advancements in technology are creating incredible amounts of wealth concentrated in the hands of a few. It’s also decoupling job creation from wealth creation.
These temporary workers or “gig economy” workers I mentioned earlier will not be the only workers affected. Traditional white collar jobs are also threatened by design thinking and technology. At the World Economic Forum in 2016, special attention was called to advances in machine learning, AI, automation, and computer vision.
As these technologies develop, entry-level opportunities in law, finance, media, and medicine will become increasingly scarce and increasingly threatened.
It is a miracle of human achievement that automated cars and trucks will soon drive alongside human controlled vehicles. The necessary advancements in computer vision and machine learning have been staggering. We must remember though, that the same technologies that must be developed to allow robotic cars to exist will begin to threaten a great many jobs when applied beyond transport.
Earlier this week, at ADA University here in Baku, President Obasanjo highlighted an example of a Nigerian with a PhD driving a truck in Nigeria because it was the only job he could find. We all responded, sadly shaking our heads considering the sad plight of this highly educated man. However…it is very possible that within ten years, that truck driving job may be automated, and that PhD in Nigeria won’t have a job at all.
President Obasanjo also highlighted the links between unemployment, lack of economic opportunity, and Boko Haram’s successful terrorist recruitment efforts. Boko Haram has become very attractive to some Nigerian men without other employment opportunities.
You may have heard, that the United States is a very litigious society. I am confident that plaintiff’s attorneys will keep robotic cars off the roads of America for many years. But, if you can imagine a nation which must move significant amounts of goods over roads and if you can imagine that country also easily removing legal obstacles and liabilities, then that is the country where you will begin to see robot controlled trucking operating on public roads within 5 years.
That country is China.
It’s not hard to imagine an overland route from China to Europe driven by robots.
During our earlier session, we learned that the United Nation’s Special Envoy for Youth estimates that the world must create 600 million new jobs to accommodate and employee young people. In a very good month, a robust economy such as that of the United States, creates 200,000 new jobs. The scale is all wrong. As machine learning, AI, and computer vision technologies roll out, the number of new jobs that must be created will not decline. On the contrary, the amount of new jobs that must be created will increase substantially. Otherwise a generations\ of young people will miss out on developing skills and experience.
Boko Haram and ISIS owe part of the origin story to negative job creation in Nigeria and Syria. We cannot underestimate the harm that will be caused by allowing youth to waste away on the sidelines.
Professor Andrew McAfee from the Massachusetts Institute of Technology wrote recently in his book The 2nd Machine Age that machine learning will fundamentally change the way that products are developed. And that Machine Learning will disrupt everything.
Today, we heard a lot of policy makers talking about education and focusing on STEM subjects as a means of reducing unemployment. But it may not help nor matter once the computers are programming themselves and making decisions based on their interpretations of massive data sets.
When we hear the term, “Big Data” it’s not about performing regression analysis on massive data sets. That happened ten years ago.
Now…the data sets are enabling computers to learn. Storing and accessing data is surprisingly inexpensive. Advances in processing and evolution in standards have made it possible for data sets to absorb information from every connected device everywhere in the world.
When you point a machine learning application toward data sets of that size…anything could happen. The process of creative destruction may now be broken.
USER PRIVACY
I want to briefly touch on user surveillance. Many would recognize that technology is a high growth sector. Increasing public mistrust in technology is not good for the economies that are dependent on technology industries.
We learned a lot about how the NSA works from Edward Snowdon. As the NSA monitor citizens, it increases mistrust. States with technology industries should not score own goals. They should not undermine their most productive industries by sewing mistrust. In the Unites States, the combined market cap of Google and Apple..the makers of Android and iPhone is over $1 trillion dollars. When you add in Microsoft it is close to $2 trillion in market cap.
So…it is disturbing to technology professionals like myself when the NSA monitors citizens without a warrant and releases its own viruses, such as STUXNET.
If you do not know the wild story of Stuxnet, it is an ingenious virus. It travelled across the world via thumb drives but would lay dormant until it recognized that the virus was physically loaded onto a machine in Iran. The virus had only one purpose, to destroy centrifuges manufactured by Siemens used in the nuclear fuel creation process.
The US Government is not alone in creating cyber attacks. China maintains a military division called PLA UNIT 61398 whose sole function is to hack NATO governments and foreign companies.
In Russia, the New York Times has reported on the existence of an office in St. Petersburg. In the office are a team of specialist propagandists whose purpose is to corrupt information and spread messages aligned with Russia’s foreign policy establishment.
A consequence of this type of government activity is the hack of the Sony Corporation. A private company was brought to its knees by state-sponsored hackers.
As a result of all this mistrust…citizens and entrepreneurs are taking extreme measures creating messaging apps that are un-hackable and which are perfectly hidden. Based in San Francisco, a company called Wickr has made an app that preserves User Privacy perfectly. Messaging is entirely secret. As you can imagine, it has become a favorite tool of organized crime and ISIS. So…American citizens who have been un-nerved by NSA and other government surveillance have created a tool to protect themselves. That tool and others applications just like Wickr are now a key component of the command and control structure of terrorist groups all over the world.
The largest American company by market cap is Apple. What should Apple’s role be when we consider the need to protect users while also considering law enforcement’s need to protect citizens more broadly? Apple’s reluctance to work with the FBI is very understandable in this climate. Their refusal is a small effort to restore trust in technology. But it does come at a cost.
I do not want to take too much of your time. So…thank you very much for this opportunity to speak with you today.
If I were to make a recommendation for you to take home to your colleagues focused on education policy, I am inclined to recommend what Governor Weld suggested this week at ADA University.
Rather than focus exclusively on STEM subjects, Governor Weld recommended that students study the classics and other disciplines that focus on developing a student’s quality of thinking.
I think that he said this because the only thing to stop technology development from doing irreparable harm is our set of values. The only thing to stop the spread of disinformation is the ability of our populace to think critically.
We can choose not to make things. We can choose not to create perverse incentives. We can choose to make the applications of certain technologies highly regulated.
I tend to agree with Governor Weld. Do not organize your schools and education systems to produce only engineers. Organize your schools to produce smart citizens capable of critical analysis, some of whom may become engineers, some may do other things. Above all make certain that your schools are educating your girls and are producing women with tremendous capabilities.
Thank you very much.