How the AI Revolution Creates New Jobs

Artificial intelligence is revolutionizing the global economy and as it progresses, it will restructure the workplace as profoundly as the industrial revolution did. To understand how, let’s dig into the change currently underway:

  • AI neural networks, using open and scalable techniques, are quickly learning (unreasonably so) to solve problems and accomplish tasks far better, faster, and more efficiently than the hundreds of millions of human beings who are currently being paid to do so. Furthermore, AIs do this in a way that eliminates most of the engineering and programming effort currently used to solve similar problems.
  • All of the major technology companies, from Google to Facebook to Uber to Tesla to Amazon are already AI companies. Some provide open tools and services for building AIs. All are building AI service platforms with better than human skill in image/facial recognition, voice recognition/speech, translation, reading comprehension, conversation, driving, flying, and much more.
  • Billions of dollars are being poured into thousands of AI-fueled startups in healthcare, transportation, logistics, legal services, marketing/sales, education, and much more. The opportunity is so great and the AI techniques so universally applicable, it’s safe to conclude that AI will be a mandatory part of every new technology start-up within the next two years. It’s also safe to conclude that there won’t be a sector of economy untouched by AI.

Where will the New Jobs Be Found?

Based on what we’ve seen so far, it’s clear that hundreds of millions of traditional jobs in all sectors of the global economy, at every income level will be lost to AIs. That’s inevitable and apparent. The hard part (the trillion dollar insight) is in understanding what lies beyond this change. This should help. It’s possible, if you look closely at what is going on already, to discern the way the AI revolution will create jobs. Here’s how:

  • Gathering Data. AIs require lots of data in order to bootstrap their learning. Most of the big companies currently use customer data. For example, Facebook uses uploaded pictures from user accounts to train its facial recognition software (it can now find you out of more than 800 million people in five seconds). Other companies use low cost workers to gather the data: Uber uses over 700 thousand “partners” to gather data needed to build a global delivery service that it hopes to automate with AI driven cars. As AIs spread into every nook and cranny of the economy, specialized data gathering will become an increasingly valuable activity, turning it into a source of large scale employment.
  • Training and Coaching. AIs can learn on their own, but they can only become very good at what they do (and stay that way despite a rapidly changing world) through human training and coaching. Many companies already taking advantage of this. For example, Tesla customers are logging a million miles a day actively training the driving AI in their cars and Google’s Deepmind is using human trainers to periodically reward and redirect AIs to radically speed up performance gains in achieving stated goals and maintain that performance despite changing real world conditions. As specialized and customized AIs become common, millions of capable human trainers will be needed to actively train and coach AIs to reach and maintain performance standards — from customer service to sales to manufacturing to law to medicine to retail
  • AI Tinkering. The tools for building AIs are already accessible and getting more so by the day. This lowers the bar for creative applications of AI. Even the early version of Google’s AI building tool (TensorFlow), was sufficiently accessible to launch thousands of AI projects were inside the company. Since then, Google has radically improved it with new training tools, image recognition, and the ability to run pre-trained AIs on Android. This advancements make it easier for a large and growing number of people, inside and outside of corporate world, to build specialized AIs that solve problems and accelerate performance. The massive value this will create means that AI tinkering will become a very valuable full time job for millions.

What Does This Mean?

The AI revolution will have a multitude of unexpected consequences. There are simply too many for me to fully list here. Here are a few of them, from the perspective of public policy:

  • Who owns the data? Some of the most valuable AIs in the world are being trained by Facebook, Google, Amazon and others based on vast amounts of customer data that is provided without meaningful compensation. If these AIs become central to the next economy, shrinking the value of the industrial economy in the process, this may become a source of social discord. The big tech company with a powerful suite of globally valuable AIs that solves this data contribution problem by turning customers into part owners could become an economy unto itself.
  • (Mechanical) Turking. The gathering of specialized data on how specific tasks are accomplished could be commoditized and globalized in negative ways. For example, as with Uber, tasks would be provided to workers via smartphone. Workers would accomplish those tasks, thereby training the new AI. Once the AI was trained, they would compete for another training role. Jobs like this could be very low paid and potentially gruelling.
  • AI Cooperatives. AIs produced by communities of AI tinkerers could emerge to challenge the proprietary AIs of the corporate world. These communities, like open source software, would be driven by vast networks of contributors actively improving AIs. However, unlike open source software, they could be the seed for the establishment of a parallel economy based on sharing streams of royalties from licensing these AIs to the corporate world and zero dollar peering relationships with other cooperatives to share AI collections w/o royalties.

John Robb

john @ johnrobb.org

This article is a framework. It’s a combination of systems thinking, traditional analysis, and broad operational experience. Frameworks provide decision makers with a tool for overcoming high levels of uncertainty. This is what I do for a living.