We see news about AI everywhere; sometimes, we see the excitement around AI and sometimes we see articles that talk about how AI will replace or destroy our jobs. We also see the occasional article talking about how AI will destroy humanity.
In this article, I will not discuss an artificial general intelligence or an evil AI that wants to destroy humanity. I will focus on current AI, which is mostly based on the algorithms that can do predictions, and discuss how the economics of AI works and how it may affect business. I also want to mention that the content of this article is highly affected by (and this author highly recommends for further reading) Prediction Machines: The Simple Economics of Artificial Intelligence and Human + Machine: Reimagining Work in the Age of AI.
This article is divided into three main parts:
- In The Evolution of Technology, I will briefly discuss the past and its similarities to the Age of AI.
- In Strategy, I will discuss how having a higher accuracy of prediction may affect strategies and business models.
- In Human and AI Interaction, I will discuss how humans can complement AI and how AI can complement human efforts.
The Evolution of Technology
Before going forward, I would like to discuss the similarities between some historical events that are analogous to how we think of AI today. I will give some examples of how a wide usage of particular technology changed our mindset. How did we get from basic arithmetic to specialized artificial intelligence development companies?
The Age of Electronic Computers
What computers do best is arithmetic. Before the computers we know now, the term “computer” was used for people who do literal computing, which we call it now a “human computer.”
With the advances of the technology, computing became cheaper and faster, and we started thinking of everything in terms of arithmetic. Photography is a good example—historically, modifying or applying visual effects to photos was a chemical reaction. Now, however, we use algorithms accessible to artists and photographers through software suites to mathematically apply effects to photographs.
This is an excellent example of how we think when the cost of a good/service drops; we start thinking about how to solve our current problems in terms of this new tech. It is the same for AI.
The Age of Internet
When the internet came to be widely used, it made huge movements in various industries, and it was all about the reduction of cost in different areas. For example, the cost of distributing goods and services became cheaper, and this triggered the birth of the eCommerce industry. Companies, eventually, changed their strategies and either survived or died.
Once the cost of a good or service drops, we start using it more often, and we can see this for the web as well. This also changes our mindset and we move whole industries online. On top of eCommerce, another example can be seen in the use of search engines; we no longer use encyclopedias to search for information but instead use Google or other search engines.
The Age of AI
The cost of AI is getting cheaper in terms of computation power and in terms of tools. Each new tool/library is helping machine learning developers to spend less time on prediction problems. For example, Google’s TensorFlow, AutoML, or even scikit can be shown as examples for this purpose. We can also show the increased usage of GPU computing as an illustration of the cost reduction in AI.
The forecast for sales for the next quarter of a company is an obvious prediction problem, but developing an autonomous vehicle was not a prediction problem a decade ago. Cost reduction in AI is changing our way of thinking, which means we started thinking of various problems as a prediction problem. We were already using autonomous vehicles in controlled environments like factories, where the vehicle could be programmed by using
if-else programming conditions. Changing the mindset and looking to this as a prediction problem helped engineers to develop autonomous cars, which can be used in the wild.
Here is how it worked, basically; an engineer taught an AI what a human would do in various conditions, and this enabled the generation of onboard software that allows drivers to use cars for thousands of miles instead of tiring out after a few hundred. AI learned what a human would do and started predicting what it should do. This is a very good example of thinking about a problem in terms of prediction.
Here’s a major question: Will AI affect companies’ strategy and business models? If you think of AI as a prediction tool which helps you to make some decision, it may not clear how it will affect pure strategy, because it is just another tool helps you to make decisions. But, if you start to think of AI as a prediction tool that can forecast with high accuracy, that may change the strategies themselves. There is an excellent example in the book Prediction Machines: The Simple Economics of Artificial Intelligence.
When we shop and purchase goods from Amazon, it ships the packages to our office/home. So, this method can be called the shopping-then-shipping method. We also know that Amazon has a recommendation engine and it recommends items while you navigate the pages. We don’t buy all the recommended items, but it at least recommends the items that we might be interested in. Let’s assume that Amazon started predicting what you will buy with high accuracy. If you started buying 80% of the recommended items, Amazon may decide to send the items before even you buy them—let’s call this shipping-then-shopping. This is an obvious change in business strategy because, once the items arrive at your home, you will send 20% of the items back, and current Amazon price modeling is not based on this assumption. Maybe, Amazon will decide to send a truck to your city once a week to collect the returned items, and this will completely change how Amazon charges your credit card, how it packages the items, and how it handles the returned items. All this strategy change is the benefit of artificial intelligence, which has higher prediction accuracy.
I believe we can work on more thought experiments like the previous Amazon example by just thinking about what would happen if AI could predict with higher accuracy.
Human and AI Interaction
How will human and AI interaction evolve in the future? Will they compete, or will they work together? I will focus on those questions by looking through the book Human + Machine: Reimagining Work in the Age of AI. According to the authors, there will be scenarios where humans complement AI and where AI will complement humans.
Humans Complementing AI
Humans can complement AI in three areas: training, explaining, and sustaining.
AI needs data to learn, which is called the training phase, so it can make predictions.
In the future, we may have training agents which are focused exclusively on training AI based on the requirements of that business. If it’s a factory, a training agent could be responsible for training a robot; if it’s an eCommerce business, a training agent could be responsible for aggregating historical data.
We need to understand how and why AI provided a specific answer to a specific problem.
Generally, we face a tradeoff between the explainability and the accuracy of AI. The black-box AI methods have more accuracy compared to methods that can be easily explained. Even though there are tools that are developed to explain why a black-box AI made a specific prediction, we may need a job role which can understand and explain the outcomes of AI.
We need to be sure the AI is functioning as expected.
In 2015, a robot in a Volkswagen factory grabbed a worker and fatally crushed him. We may need roles whose responsibility is to ensure that AI systems are working as expected.
AI Complements Humans
The potential of AI gives humans superpowers because AI makes predictions faster and more precisely than humans ever could. These superpowers can be expressed in the value that they bring to a given situation or action.
AI tools help humans to increase the capabilities of being human. In the book Human + Machine: Reimagining Work in the Age of AI, the authors use the example of Autodesk’s Dreamcatcher software, which uses genetic algorithms to iterate through possible designs.
A designer can design a chair light, cheap, and strong with the help of this tool. AI tries to create a design which is based on the given criteria and provides the results to the designer. The designer then uses chooses one of the given designs and uses their creativity on that design to make the final touches.
This is similar to the what computers provided to people with the computer age—just at a new and exciting level of capacity in terms of what kinds of things AI can assist with.
AI can act as an assistant to help people by interacting with them. Amazon’s Alexa, Google Home, and Apple’s Siri are prominent examples of this kind of interactive AI agent. As those agents are improved with each iteration, we will start using them more often, and it will become a part of us, like a deeper version of what we are doing with our smartphones. Those agents will be our private assistants and they will complement us.
Examples of AI-fueled physical augmentation can be found in factories. Although factories are operated by robots right now, they are mostly rule-based systems and put in a cage—just in case—for safety. Robots will help humans as co-workers and will be designed not to harm people while freely moving and working in a factory.
Although there are some concerns that say “robots are more efficient, so human workers will be discarded in future,” Markus Schaefer, head of production planning at Mercedes, says, “We’re moving away from trying to maximize automation, with people taking a bigger part in industrial processes again.” New technologies do bring monumental shifts in the way we do things, but the invention of the plow did not eliminate the need for farm workers, nor did the invention of the computer eliminate the need for mathematicians. As with all technological revolutions, the advent of AI will be utilized to help humankind reach a new paradigm, not to replace it entirely.
Ready for more technical AI knowledge? Try A Deep Dive into Reinforcement Learning to learn how to teach an AI to drive a car up a mountain.
Understanding the basics
The applications of artificial intelligence vary from autonomous cars to translation, from chatbots to image recognition. Digital assistants like Siri and Alexa are typical examples of AI applications, and with recent increased efficiencies in AI, we should see more AI applications in the future.
Basically, you can use AI in two different areas in your business; you can build your business on AI, which means your core value proposition will be a technology which is based on AI. The second method is leveraging AI methods to improve business and increase the volume of sales.
Artificial Intelligence was a term coined by John McCarthy in 1955 and is a field of computer science that studies machines (algorithms) that work and react like humans. AI has sub-fields like machine learning, natural language processing, and computer vision.
Predictive algorithms predict an outcome based on given inputs. They operate in either a learning or prediction phase. In the learning phase, historical data of input-output pairs is passed to the algorithm and it maps a relationship between the two. In the latter, only inputs are given and it predicts an output.