Will Reinforcement Learning be used in the workplace?


A joule is the SI base unit for energy. In physical terms, lifting an apple one meter from the ground takes about 1 joule of energy.  If you were to pay someone to lift that apple at minimum wage, it would cost an employer $8.30 an hour (Minimum wage in Ohio 2018). [2] Therefore, an average workday will cost the employer $66.40. Moreover, humans need to take breaks often and are not able to perform such repetitive motions consistently and safely. In contrast, a machine lifting that apple will cost $0.12 per kilowatt-hour (Average cost of electricity in Ohio) or $0.00000003 per apple lifted one meter. This is significantly cheaper then employing a human to perform the same task; however, there is upfront cost with the machine itself.  Also, not all tasks are as simple as lifting an apple. Most jobs are multidimensional and complex. Machines need to have a complex algorithm that can perform tasks that a human would be able to do on a normal work day. This is where artificial intelligence is showing potential today and forecasted to overtake humans in terms of performance.

Contemporary Technical Solutions

Artificial intelligence is a branch of computer science that can display intelligence with software. Some of these abilities include perception, searching, planning, and learning. In industry we need robots that can form complex tasks and find optimal solutions. Reinforced learning is an area of AI that focuses on how an agent might act under an environment in order to maximize some given reward.  This can be illustrated as a cleaning robot getting rewarded as it navigates its environment successfully while picking up trash from the ground. The agent only needs a reward for accomplishing a task to find the optimal policy for a complex system.

Reinforced learning can be thought of as computational behaviorism. This is very similar to human and animal behavior, insinuating the interaction with the environment is the defining factor of behavior. This algorithm sits at the intersection of many different fields of science, and has many applicable examples. In engineering, there is a type of problem called optimal control. This reward system is based on optimally deciding on a sequence of actions to get the best results. Furthermore, in neuroscience the reward system of the human brain is studied, and is based on how we execute actions and generate dopamine. Another instance of reinforced learning occurs in mathematics, there is field called operations research that deals with the application of advanced analytical methods to help make better decisions. Overall, the most promising case of reinforcement learning is its use in video games, and it is starting to make its way in robotics.

The main premise of reinforced learning is its ability to generalize large environments and find optimal policies to accomplish its tasks without any human intervention.  Some issues with artificial intelligence is the fact we need to train the algorithm with supervised means. This means the algorithms can only learn as good as the human teaching it. This has been a problem with some environments like board games. The Japanese board game Go, is one such example because there are more board configurations in the game then there are atoms in the known universe. Obviously, we can’t write an algorithm to recursively know the best answer because of computational problems with today’s computers. We can teach it using the best human players, but it won’t be any better than the top performers. The only successful algorithms produced thus far were created by researchers at Google Deepmind. Using reinforcement learning techniques, scientists were able to outperform the top players in the world only after exposing an agent and letting it play itself for a few days, insinuating the best teachers for such technology are the programs themselves.[7]


Today the most powerful companies in the world are making large investments to artificial intelligence research. Big tech companies such as Google, Microsoft, Amazon, Apple, and Facebook are all investing in AI and are highly profitable because of this new technology. We have more data available to us than ever before in history and are finally starting to make use of it. We are currently witnessing AI perform incredible feats and it’s not likely to stop. Regardless of what happens in the next few years, artificial intelligence is here for the future. For now, AI has its place in industry but for many it’s a growing philosophical need to understand ourselves and how we think.