What do we know about AI?
In our previous article, we’ve already discussed AI, namely, the definition of AI and the scope of its application. Today, PaySpace Magazine would like to continue the series of articles about AI, but the topic will differ slightly. We’ll consider what AI means for scientists, and try to find out what we really know about AI.
In 2020, we are facing a new issue. Most managers and marketers call artificial intelligence everything they want to call: vacuum cleaners, toy transformers, and even the option to select special-offer mobile tariffs. AI has become a trend, and is selling well. However, there is one issue. Even scientists do not dare to say that they have created AI.
Even though we’ve published an article about the general definition of AI, we have decided to find out whether we are talking about real artificial intelligence, how it differs from machine learning from the point of view of scientists, and is it fair to contemptuously mock another advertisement of the “newest-most-advanced” device/service powered by AI.
AI and actual intelligence
In our previous article about AI, we’ve tried to provide the most comprehensive and universal definition of AI. Some scientists believe that the definition of AI is not just universal, but rather abstract.
First of all, let’s figure the definition of AI. For example, Wikipedia says:
Intelligence is the capacity for logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning, creativity, critical thinking, and problem-solving. More generally, it can be described as the ability to perceive or infer information, and to retain it as knowledge to be applied towards adaptive behaviours within an environment or context.
Yes, the definition of intelligence seems to be quite universal and abstract as well, but you also should know that intelligence is considered to be the property of the mind (referred to as the psyche), which can be possible for non-humans (animals, and even plants), but still, living organisms. Another intelligence definition says that it is about the ability to adapt to new situations, the ability to learn and remember, based on experience gained, understanding and applying abstract concepts, and using one’s knowledge to manage a person’s (a living organism’s) environment.
Frankly, it didn’t clarify the issue for me. Therefore, let’s see what scientists and AI labs think about it.
AI and science
Globally, the AI question is not simple, because even specialists worldwide still can’t agree on one universally accepted unified definition.
Therefore, we have decided to take a deeper look at what labs and AI departments are studying. Mostly, AI scientists are studying:
- Predictive analytics, which is essentially data mining. Algorithms can make a forecast based on this data mining. For example, some banks use similar systems to decide whether they can grant a loan. Such technology can even analyze the customer’s social networks, and construct a psychological profile with subsequent evaluation of the reliability of the applicant.
- Recommender systems, which are the algorithms that select and offer objects for the users, based on their previous choices, i.e. content, products, services, etc. We are all familiar with these phenomena. Let’s say, you feel blue in the middle of the night, so you decide to play the Titanic soundtrack on YouTube. Be sure, the system will remember you as a Celine Dion fan, or even worse, the “sad playlist for the feeling suicidal sadness” appreciator.
- Computer vision and image recognition, which are AI fields that teach computers how to interpret and “understand” the visual world. The best examples of the practical use of these technologies are unmanned vehicles, or in services like “FindFace”, “SearchFace” “FindClone”, which can search the photos that match the face you search (uploading a photograph, recognizing a face in the image, and then matching this face to its “twins” on social media).
- Synthesis, recognition, and generation of spoken language, and this is what we call “virtual assistants” use (i.e. Siri, Bixby, Alexa, etc).
Definitely, the term “artificial intelligence” is used in tasks, where the system analyses data, and makes so-called “smart” decisions on the basis of this.
But note, there is no researcher that claimed that he created AI. As of today, scientists can only study algorithms that perform tasks from the field of AI.
AI and machine learning
AI is often confused with machine learning, but that’s not entirely true. Machine learning (ML) is often used for similar tasks, because it is convenient to analyze data and make a decision with its help. For example, the ML-algorithm can say that the person from the picture is indeed a smartphone user (with a probability of 98%). Thus, a phone can be unlocked.
But scientists do not equate ML with AI. For them, artificial intelligence is a field of research on how to make a machine to perform non-trivial tasks. ML, in its turn, is a class of algorithms that serve to solve issues.
AI and elitist mindset
Those who like to get to the bottom of the matter used to say: “This vacuum cleaner is not smart because it cannot cook pizza or discuss Nietzsche. AI should be like an oracle that is ready to answer any question and solve any problem.”
Such an idealistic image of a super-machine is instilled mostly by cinema and art. The chatbot that fails to operate after the first request looks poor compared to some super-robot from another big-budget science fiction film.
There is a special definition, which fully describes what such snobs mean by artificial intelligence: strong artificial intelligence, or artificial general intelligence. Basically, it is a utopian algorithm that will cope with any task unconditionally, like an infallible hero.
One of the theories of strong AI says that computers can acquire the capacity to think and become self-aware beings (in particular, to understand their own thoughts).
On the other hand, there is another term that derives from the abovementioned definition, and this is an applied AI (also called weak or narrow AI) we are talking about. This type of AI is created to perform specific tasks. For example, one algorithm in a drone analyses the road, while the other one, based on this data, understands where it should go. Yes, this is ML, but nevertheless, it is still the field of AI. But unlike artificial general intelligence, it will not be able to act in conditions of uncertainty, meaning it will not be able to raise kids, or, let’s say, save the world.
The bottom line
Marketers and advertisers label all the newest smart devices that can do something by themselves as “AI”. But a device that is able to drive around the apartment, turn on the lights, or pick up the ordered parcel is not “AI”.
As of today, the human race is far from creating a real artificial intelligence, maybe, when we are talking about things somehow related to AI, it would be right to use the definition like “machine learning algorithms” or say that the technology is built on AI algorithms. Indeed, putting a “smart” vacuum cleaner on a par with the idea of AI is at the very least unfair.