As an academic discipline, Artificial Intelligence has been around since the 1950s and has gone through several “summers” and “winters” during this time. In the past decade AI started to enter businesses on a new level of attention, investment readiness and productivity. Most experts argue that we are on a breakthrough point regarding mass adoption of the technology.
This hypothesis can be supported by the massive success of companies using Artificial Intelligence as integral part of their products and services. Amazon, Apple, Facebook, Alibaba, Tencent, Microsoft and Alphabet — all belong to the league of the most valuable and industry shaping companies — are betting heavily on the technology. Let’s get down to the why, what and how of Artificial Intelligence.
Our Definition of AI
“Artificial Intelligence (AI) is the ability of a computer-controlled entity to perform cognitive tasks and react flexibly to its environment in order to maximize the probability of achieving a particular goal.
The system can learn from experience data, and can mimic behaviors associated to humans, but does therefore not necessarily use methods that are biologically observable.”
This definition is a blend from various renown sources, aiming to cover the most important aspects without being too vague or depending on very accurate further definitions of the terms “intelligence” or “machine”. It might also have its shortcomings, but the point is, that there is no perfect definition for AI, yet. There might never be one, if technology continues to evolve that fast.
The important part to understand is, that there are two different approaches to build an Artificial Intelligence: The first, the Artificial General Intelligence (AGI), often associated with a human-threatening dystopia and the second, so called narrow AI, which is highly specialized to perform a very particular task on the level of human capabilities or even beyond.
While the first one is a future scenario and requires rather a philosophical approach to be discussed in depth, the latter one is already augmenting many tasks in businesses or even in your smartphone. Interestingly, as AI gets more important and productive in the daily life, people tend to remove things like Siri, person recognition in your iPhone photo album or product recommendations on Amazon from the definition.
AI classification — another attempt…
AI is a collective term of various subfields. Oftentimes, applications, tools and functions are somehow confusingly aggregated. Let’s try to find a common basis here. On the inner circle, there are six major cognitive functions mimicked by AI:
Artificial Intelligence domains
- Machine learning
- Knowledge representation and reasoning
- Automated Planning
- Natural language processing
- Machine Perception
- Intelligent Robots
The second layer shows (almost certainly not exhaustive) corresponding sub-domains, in which the major six categories can be functionally categorized. But it is getting messy already, because some disciplines interact very closely like for example perception and natural language processing in many areas of applications (translation or your Amazon Echo speaker).
Okay, we know so far, that AI is about computers achieving certain goals, there are narrow AIs that show already today tremendous business value and these specialized AIs work in at least one subset of the before mentioned framework. Next up: underlying techniques, areas of applications, success factors and limitations.
How does it work?
Let’s not overdo it here, but rather have a very high-level look at the fundamental techniques that the computer system can use to achieve goals. There are countless theories and variations of statistical approaches to solve problems. It is almost impossible to visualize the possibilities of combining different techniques in certain areas of application. On the most general layer, you could mention:
- Artificial neural networks (inspired by the functionality of biological brains)
- Mathematical optimization (for deterministic problems)
- Probabilistic methods (for problems involving uncertainty)
- Heuristics (as trial and error method)
Especially neural networks gained traction through the improvements in Deep Learning (a sub-sub-domain of Machine Learning) in the last years. Game playing neural networks gained media attention (AlphaGo) and especially decision making and pattern recognition was highly improved through deep learning neural networks and Reinforcement Learning (a reward-driven technique).
As a consequence of this development, many areas of application are on the rise today.
Relevance of AI domains in selected industry applications
In the table you find a few use cases of multi-billion dollar industries, where AI is not only part of the corporate strategies, but also offers significant improvements through either additional revenue or cost-savings. As you see, Machine Learning plays a significant role here. By using AI it is possible to augment products and processes, improve quality, speed and processual overhead in almost all industries.
IDC estimates global spending on AI systems to reach $57.6 billion in 2021, with an annual average growth rate of 50% between 2016 and 2021. Analysts from Gartner look at the output side and project that AI augmentation will generate $2.9 trillion in business value by 2021. This would equal a global return-on-investment of over 50X. Wow! But despite maybe arguably projected numbers, the massive business impact of the technology is something almost all experts agree on.
Andrew Ng, one of the most respected experts in the AI field, shares his thoughts on the business impact of the technology in this short interview:
How to be successful in AI?
There are four major ingredients, if you want to seriously start or transform business processes on the basis of Artificial Intelligence:
- Data (!)
- Computational power
Most importantly you need data. A lot is good, but high quality data is what actually counts to feed your intelligent machine. If you don’t have good data, better start getting it! With Social Media, Internet of Things and personal smart devices such as phones, the amount of global available data has increased significantly in the past, which is one of the major drivers of the current AI hype.
The second important aspect is talent. Find very smart people, who are able to compose solutions and set up frameworks that actually provide results that can be immediately used for your business purpose. Data scientists and programmers are a scarce resource at the moment, so do not expect to hire them on a low budget.
Third, you need algorithms, that actually solve your problems. Most likely you will get algorithms in the near future like WordPress templates or e-commerce shop systems — basically ready to use — you just need to “configure” it to your preferences through tailored goal functions or training data.
Last but not least you need computing power to process your data. If it is supersensitive data, you might need internal solutions — for most cases cloud providers will be happy to provide their services. Sufficiently available, scalable and affordable computing power is another major driver of the AI development in the recent years.
Finally, a bit of expectation management. Yes, Artificial Intelligence is one of the most powerful technologies as of today. But it is not wizardry, it develops over time and it will give you useless results, if you:
- Have inaccurate input data
- Do not set a narrow focus on one specific task at a time
- Expect that self-learning immediately equals autonomous systems
- Do not create organizational understanding and acceptance for the technology
References and further definitions of AI
AI is typically defined as the ability of a machine to perform cognitive functions we associate with human minds, such as perceiving, reasoning, learning, interacting with the environment, problem solving, and even exercising creativity. Source
Artificial intelligence (AI) refers to systems that show intelligent behaviour: by analysing their environment they can perform various tasks with some degree of autonomy to achieve specific goals. Source
It is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable. Source
Artificial intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals. Source