A familiar ring echoes throughout your room. It’s the CEO of your company Hal, calling for you to come at the corporate headquarters today.
Shortly after arriving and sitting in your workstation, you are greeted by Hal, who opens with a few familiar words and a set of text, on what looks like a VOIP and a chat session. Turns out, the voice wasn’t from a human at all.
Hal is actually an AI: a comprehensive management system that is made to establish, build, maintain and supervise a corporate organization without any direct human agency. In other words, your boss is a robot CEO.
What would it be like to work for a robot CEO? Come along with our time machine of the imagination, as we look at how today's technology might develop over the next few decades. Is the future likely to be chock full of synthetic managers? We shall see!
Discussing with Hal
With the greetings out of the way, Hal starts to relay the important business of the day over to you. It provides detailed information via chat and text, only using the tables and data as supplementary materials. Demeanor-wise, Hal probably sounds a bit cold.
Yet, Hal still sounds convincingly human. Its voice and speech has this controlled degree of liveliness completely incomparable with Siri or Cortana of the old yesteryears.
Being a CEO is being a leader. This means interaction must be one of the primary responsibilities of Hal. It must have the tone, the cadence, and the grammar to be able to speak naturally to a real person. Now, it may sound like the stuff of science fiction, but take a short listen to this particular segment of Google’s IO 2018 event.
If you clicked that link then you'll have just listened to an updated feature of Google Assistant, named 'Duplex'. Far from the monotonic speech synthesis programs of Cold War era number stations, it speaks rather fluently. In fact, it probably convinced the people on the other side of the conversation that they are actually speaking to a human person.
If the technology is here right now, we can simply extrapolate down the road what it could be like in the next few decades. With Duplex, it is probably safe to assume that it won’t be long before robot CEOs like Hal interact with its employees much like a real corporate bigwig would.
Thinking with Hal
As the discussion moves forward, concepts and ideas start to flow. Not full ideas, but blurbs, and tidbits. Hal demonstrates an uncanny level of professional familiarity with the subjects. It then goes to adjust and adapt to your suggestions.
Not all are accepted of course, but enough thought goes through for your robot CEO to “think and contemplate”.
A robot CEO needs to be able to process and analyze information given to it using standard modes of communication. It has to learn from it so it could integrate the acquired data to its overall decision making algorithms.
Geoffrey Hinton, the 'Godfather' of Deep Learning, was enamored with the idea that to understand the brain, one must recreate the brain and how it works. As we have briefly discussed before, deep learning systems provide this very mode of acquired data, mimicking the human brain by creating a web of information from an initial set of data.
IBM’s Watson brilliantly demonstrated this ability to “think and contemplate” when it crushed its human opponents in the game show 'Jeopardy', back in 2011.
This particular show - as you may know - is not built around a traditional quiz setting. Instead, it is the players that generate the questions from a given predicted answer. Players classify an information blurb, and generate new sets of data (in this case, a question) related to it.
Future robot CEOs would not only benefit greatly from deep learning systems, they actually need to be based on one. In the near future, the birth of deep learning systems that could supersede the likes of Watson, DeepMind and Magenta, could make management AI robust enough to actually challenge the role of the corporate top brass.
Managing with Hal
After much contemplation, ideas and concepts then coalesce into new plans and strategies. Hal decides to modify portions of the company’s daily business operations in accordance to tactically anticipated changes in market trends.
However, it does not consult a board of analysts to assess its own decision. Even stranger, some of its actions are quite peculiar, almost unrelated to the decided end-goal.
Predictive software had its roots as early as 1952, when UNIVAC successfully predicted the unlikely victory of Dwight Eisenhower in the elections. It was able to analyze the given data in a way that was not as apparent with its human counterparts.
Later in the 1980s, systems such as NYSE’s SuperDOT allowed the computerization of order flow in financial markets. This gave rise to electronic trading, which then developed into algorithmic trading, literally transforming the way stock markets operate today.
On the engineering side, computer simulations were able to predict meteorological patterns, biological growth, trajectories of space probes, and other information that requires huge sets of data to work with. In these fields too, predictive software grew and developed, with the associated hardware trailing very close along.
Hal won’t be designed exactly the same way as these pioneering applications. Though, a significant part of its system design philosophy is expected to work similarly. Its own predictive algorithms should be able to see underlying patterns around even larger sets of big data, in a way not only completely alien to our limited human perspective, but much more advanced than the best market prediction software currently on the planet.
Working with Hal
Finally, Hal entrusts you to supervise for the rest of the day the one thing that it cannot manage on its own: hardware maintenance.
As the employee professionally hired for this task, there is no other human in this firm that can do this job better than you.
In fact, there is no other human in this firm!
A CEO’s job may be to lead and empower employees. But in the near future world of automation, robots, and artificial intelligence, humans may not need to apply anymore. The demand for information-driven human labor, may it be for low skill or even professional jobs, could soon disappear.
In the United States alone, it is predicted that from the years 2016-2025, revenue generated by artificial intelligence will grow up to 9 billion US dollars. During the same time period, almost 47% of all jobs could be taken over by AI. All of that, in under a decade.
If something as competent as an actual artificial intelligence CEO could ever exist in the next decades after, then it is logical to assume that at that time, the robots have financially taken over the world.
Photo by Franck V. on Unsplash.