What is artificial intelligence?
Artificial intelligence is a branch of computer science that deals with automating intelligent behavior. The explanations of terms for artificial intelligence are as varied as the attempts to explain the technology in simple terms. A generally valid definition does not yet exist. To the question: “What is artificial intelligence?” However, to be able to answer in an easy to understand way, there are certainly a few delimitations possible, which we will look at below.
The ideas for today's artificial intelligence come from applied information technology and the field of neuroscience. At first glance, this appears to be a contradiction — after all, information technologies and the science of the structure and functioning of nervous systems do not have much in common. And yet there is this small intersection, which involves researching mechanisms of intelligent human behavior. For example, many current computer programs and dialog systems are like self-learning chatbots inspired by their natural role models and show intelligent, i.e. learning, problem-solving behavior in their algorithms. In a few exceptions, they even exceed the results of human intelligence.
Back in the 1950s, Allen Newell built up a first model of human thinking based on GPS (General Problem Solver), in which — depending on the context and state — the program can learn and decide independently.
Artificial intelligence is therefore...
a system of algorithms that learns to solve problems based on (sometimes very large amounts of) input data. The input data can be examples that have already been solved here, or simply the texts from many websites or the steering movements of a car driver together with the (simultaneous) images of the road. Training artificial intelligence, which is also called “training”, makes sense especially when conventional algorithms that are “only” based on previously defined rule sets could not complete the task or could only be completed with extremely high programming and computing costs. For example, it is difficult to draw up a set of rules that instructs you to recognize handwritten text because there are so many manuscripts that differ in important nuances.
AI systems are already integrated into our everyday lives and are intended to solve problems there using innovative methods. Examples include self-learning chatbots: translation programs, automated image enhancement in cameras and smartphones, or automatic recommendation systems. How important AI is for all areas of technology can be seen not least from the fact that there are hardly any new computer/smartphone processors that are not designed to massively support/accelerate AI.
When is a machine considered intelligent?
Similar to intelligence tests in humans, there are also standardized task sets for machines/models that should “measure” their intelligence and decide whether they can join the artificial intelligences cafeteria club. Unlike human tests, however, tests for AI models are always limited to specific areas of responsibility, domains.
One of these benchmarks is GLUE, which uses 16 different subtasks to evaluate an NLP system (the domain here would be language comprehension).
These include:
- Similarity in meaning of sentences
- Sentiment analysis
- Question-answer test
NLP systems that do well in such tasks form the basis for modern chatbots. And even though there is no perfect NLP system yet, today's leaders learn something new with every application and every training.
How does artificial intelligence work?
As artificial intelligence is becoming more and more entrenched in our everyday lives, there are also more and more players, both from the corporate world and from the open source community, who are simplifying the use of AI. Learning powerful AI models in particular still requires a lot of computing power and a great deal of training data today. Some areas of AI, such as “deep learning,” adjust billions of parameters in artificial neural networks until these networks “make good decisions.” Such models, such as GPT-3 from OpenAI, manage to be mentioned even in non-tech-savvy media due to their spectacular performance.
But fortunately, numerous companies provide precisely this basis for free or with fair licensing costs. In the NLP sector, for example, Hugging Face and its community should be mentioned here, which provide “pre-trained” language models and toolboxes, which are then fine-tuned and put together accordingly by companies such as moinAI. Using these tools then requires only a fraction of the computing power and data compared to the basic models. A good cooking recipe benefits from the ingredients, of course, but it is the composition and seasoning that makes it a delicious dish that you can also serve to guests.
Areas of application of artificial intelligence
Artificial intelligence often paves the way for human decisions or relieves stressful work. There are many examples that show how diverse the areas of application for artificial intelligence are.
Autonomous driving with artificial intelligence: Keep your eyes peeled at the wheel?
Networks with many levels, the so-called concepts, also contribute to process automation or autonomous driving. Intelligent programs can solve these tasks within a very short time and learn complicated concepts.
Autonomous driving depends on artificial intelligence and requires the technology for signal processing and/or environment recognition, for example. However, there is still a long way to go before a car can move completely autonomously in every situation, i.e. without the intervention of the driver. The capabilities of a self-driving car are divided into five levels:
- Assisted driving
- Partially automated driving
- Highly automated driving
- Fully automated driving
- Autonomous driving
Level five therefore means that the car can take on full autonomy. Intelligent software controls and implements all driving tasks. Cameras, sensors and central control units generate huge amounts of data. These should answer relevant questions in real time — for example how good the current visibility or road conditions are. Artificial intelligence must therefore constantly make decisions.
Today, driver assistance systems still largely leave the initiative to the driver, but Mercedes even has official German road approval for a first system to let the car autonomously steer himself in certain situations (low speed) on the motorway, so that the driver can watch videos. In just a few years, however, many more situations should be handled fully autonomously. It remains to be seen whether this forecast will actually materialize in view of legal regulations, moral considerations and technical requirements.
Artificial intelligence in Industry 4.0
Industry also uses artificial intelligence in a variety of ways. Almost every fourth machine is now smart. AI makes it easier to deal with complex relationships and leads to networked processes. In the optimized manufacturing process, machines and components communicate with each other, for example through image analysis and image recognition. Another example from the industrial sector is collaborative robots, which not only relieve employees of physically heavy work, but can also increase assembly precision. AI enables improved quality control around the clock.
Artificial intelligence in everyday life
Smart processes and machines help us with everyday tasks. They can read, see, speak and, above all, understand through continuous learning. For example, specialized software is able to transcribe hours of audio recordings. The well-known voice assistants Alexa, Siri and Cortana have become little everyday helpers and intelligent chatbots are waiting for you in customer service with individual solutions.
- Medicine (for example for quality of care and improved diagnostics as part of imaging techniques)
- banks (for example individual investment tips from a robot advisor)
- Translator (for example, translating long sections of text within a few seconds)
- Smartphone (e.g. unlocking the device using facial recognition)
- Emails and messages (with integrated spam filter, voice assistance, etc.)
- Google search (AI improves the processing of search queries using “RankBrain”)
- Netflix and other streaming providers (for highly personalized and personalized recommendations)
- Communication with companies (e.g. with questions about products via AI chatbot)
The examples listed show what artificial intelligence is and can do. Of course, such systems do not yet work perfectly and even if they do, it always takes a human eye to rule out possible sources of danger.
However, self-learning chatbots in particular have the potential to become even more profitable in the future and to sustainably increase their quality, as they can continuously learn and take on time-consuming tasks. You can find out which tasks these may be in our free Read white papers about chatbot use cases.
How do self-learning chatbots use artificial intelligence?
Self-learning chatbots are a successful example of an effective application of artificial intelligence and also help answer the question of what artificial intelligence actually is. They show the enormous benefits of the technology for their users at different stages of the customer journey.
A self-learning chatbot with artificial intelligence registers the issue based on natural language recognition. But that is by no means all: It also records the mood of the customer at the moment. The lines between “real” human dialogue and chatbot contact are blurring.
Self-learning chatbots use every new task to expand their knowledge database. They are available and ready to use 24/7. In customer service, self-learning chatbots use AI to address customer problems, questions or complaints. A relief from which not only employees but also customers benefit: Requests are quickly received, forwarded or, in the best case, completely resolved by the chatbot itself.
Excursus: Does artificial intelligence threaten jobs?
What should artificial intelligence be able to do? This question also promptly raises concerns about the loss of jobs. However, this is completely unfounded: Intelligent programs do not replace employees, but support them.
Self-learning chatbots, for example, can take on standardized tasks and sequences and thus relieve staff — for example when contacting customers, completing inefficient tasks and pre-qualifying or answering repetitive questions. This creates new capacities for other important tasks.
Since artificial intelligence cannot completely replace humans, their behavior must also be tested and further developed by a real person over and over again. Computer programs control every artificial intelligence and artificial neural network. This software is created and programmed by humans, numerous technically oriented workplaces have already been created and demand will increase even more in the future. As we have seen, there is a lot of potential in industry, economy and services for the use of artificial intelligence in the world of work.
Dialogues with self-learning chatbots are becoming standard
Artificial intelligence will be given more and more importance in the future — this is already shown by the published strategy paper by the Federal Government. AI systems will also penetrate areas that still seem closed today.
Even a high volume of requests is no problem for an artificial neural network: Whether one or thousands of messages arrive is no problem for the network.
The question “What is artificial intelligence?” In this context, it can also be answered as follows: Artificial intelligence is like a type of digital employee who supports, relieves people and creates freedom through problem-solving behavior at their workplace.
