Not all AIs are the same. Intelligent language assistants are just as much a part of the collective term artificial intelligence as machine translation or face recognition. However, a face recognition system is not able to translate text or to understand and answer questions.
So the AIs have one thing in common: they are specialised in one concrete task, their “intelligence” is limited.
Weak ≙ Narrow ≙ Specialised ≙ Economic artificial intelligence
First of all: all existing systems can be classified as narrow artificial intelligence.
These systems were developed to fulfil a clearly defined task. Therefore, they always resort to the same methods when solving problems. They are purely reactive, do not gain a deep understanding of the actual problem or have their own independent consciousness. Humans can conduct interviews, write poems and learn different languages – machines can either do one thing, the other, or (in most cases) neither.
Also, some of the systems that are classified as such weak artificial intelligence are – contrary to any intelligence – purely rule-based. For instance, automated text generation with natural language generation works through a template-based process: while a grammar-based approach would try to comprehensively simulate human intellect in order to be able to write texts of any kind independently, this template-based approach uses pre-formulated text parts which are automatically compiled on the basis of the stated story plot.
However, these systems are increasingly supplemented or wholly replaced by machine learning. In the case of text generation, machine learning already supports the user, e.g. through a linguistic analysis and an automated adaptation of his or her templates.
Strong ≙ General ≙ Scientific artificial intelligence
Systems that represent human intelligence in its entirety and are able to independently solve new problems instead of being limited to simulating certain human abilities are classified as strong artificial intelligence. In advertising or in public, the term “super intelligence” is often used to describe this form of AI.
This suggests that there are already systems which think and act independently, and consequently are able to plan, execute and learn. But again, such a strong artificial intelligence does not exist in reality. Also, whether it will ever exist is not conclusively clear. Although meanwhile many experts tend to think strong artificial intelligence will exist at one point, it is still unclear when.
Status Quo: Augmented Intelligence
To describe the status quo as well as the future of artificial intelligence, an interpretation of “AI” as “augmented intelligence” instead of “artificial intelligence” might be more appropriate. Human intelligence is a complex construct that is difficult to define and can’t be replicated one-to-one. So whether weak or strong, “augmented intelligence” tends to be at the core of what the technologies are ultimately about.
Just as the term robot journalism describes the use of automated text generation in the media context in a striking way, but is actually quite misleading, the question arises how accurate the term “artificial intelligence” is, and to what extent machines can be intelligent at all. Machine (intelligence) should be seen as an extension of our potential that allows us do things that we ourselves could never do – and vice versa.
Ultimately, the question is not how to simulate our human intelligence in the most efficient way, but how to expand it profitably.
About Retresco | @retresco
Founded in Berlin in 2008, Retresco has become one of the leading companies in the field of natural language processing (NLP) and machine learning. Retresco develops semantic applications in the areas of content classification, recommendation, as well as highly innovative technology for natural language generation (NLG) . Through nearly a decade of deep industry experience, Retresco helps accelerate its digital transformation, increase operational efficiencies, and enhance customer engagement.