Strong artificial intelligence (AI), also known as artificial general intelligence (AGI) or general AI, is a theoretical form of AI used to describe a certain mindset of AI development. If researchers are able to develop Strong AI, the machine would require an intelligence equal to humans; it would have a self-aware consciousness that has the ability to solve problems, learn, and plan for the future.
Strong AI aims to create intelligent machines that are indistinguishable from the human mind. But just like a child, the AI machine would have to learn through input and experiences, constantly progressing and advancing its abilities over time.
While AI researchers in both academia and private sectors are invested in the creation of artificial general intelligence (AGI), it only exists today as a theoretical concept versus a tangible reality. While some individuals, like Marvin Minsky, have been quoted as being overly optimistic in what we could accomplish in a few decades in the field of AI; others would say that Strong AI systems cannot even be developed. Until the measures of success, such as intelligence and understanding, are explicitly defined, they are correct in this belief. For now, many use the Turing test to evaluate intelligence of an AI system.
The all new enterprise studio that brings together traditional machine learning along with new generative AI capabilities powered by foundation models
Alan Turing developed the Turing Test in 1950 and discussed it in his paper, “Computing Machinery and Intelligence” (PDF, 566 KB) (link resides outside IBM). Originally known as the Imitation Game, the test evaluates if a machine’s behavior can be distinguished from a human. In this test, there is a person known as the “interrogator” who seeks to identify a difference between computer-generated output and human-generated ones through a series of questions. If the interrogator cannot reliably discern the machines from human subjects, the machine passes the test. However, if the evaluator can identify the human responses correctly, then this eliminates the machine from being categorized as intelligent.
While there are no set evaluation guidelines for the Turing Test, Turing did specify that a human evaluator will only have a 70% chance of correctly predicting a human vs computer-generated conversation after 5 minutes. The Turing Test introduced general acceptance around the idea of machine intelligence.
However, the original Turing Test only tests for one skill set — text output or chess as examples. Strong AI needs to perform a variety of tasks equally well, leading to the development of the Extended Turing Test. This test evaluates textual, visual, and auditory performance of the AI and compares it to human-generated output. This version is used in the famous Loebner Prize competition, where a human judge guesses whether the output was created by a human or a computer.
The Chinese Room Argument was created by John Searle in 1980. In his paper, he discusses the definition of understanding and thinking, asserting that computers would never be able to do this. In this excerpt from his paper, from Stanford’s website (link resides outside IBM), summarizes his argument well,
“Computation is defined purely formally or syntactically, whereas minds have actual mental or semantic contents, and we cannot get from syntactical to the semantic just by having the syntactical operations and nothing else…A system, me, for example, would not acquire an understanding of Chinese just by going through the steps of a computer program that simulated the behavior of a Chinese speaker (p.17).”
The Chinese Room Argument proposes the following scenario:
Imagine a person, who does not speak Chinese, sits in a closed room. In the room, there is a book with Chinese language rules, phrases and instructions. Another person, who is fluent in Chinese, passes notes written in Chinese into the room. With the help of the language phrasebook, the person inside the room can select the appropriate response and pass it back to the Chinese speaker.
While the person inside the room was able to provide the correct response using a language phrasebook, he or she still does not speak or understand Chinese; it was just a simulation of understanding through matching question or statements with appropriate responses. Searle argues that Strong AI would require an actual mind to have consciousness or understanding. The Chinese Room Argument illustrates the flaws in the Turing Test, demonstrating differences in definitions of artificial intelligence.
Weak AI, also known as narrow AI, focuses on performing a specific task, such as answering questions based on user input or playing chess. It can perform one type of task, but not both, whereas Strong AI can perform a variety of functions, eventually teaching itself to solve for new problems. Weak AI relies on human interference to define the parameters of its learning algorithms and to provide the relevant training data to ensure accuracy. While human input accelerates the growth phase of Strong AI, it is not required, and over time, it develops a human-like consciousness instead of simulating it, like Weak AI. Self-driving cars and virtual assistants, like Siri, are examples of Weak AI.
While there are no clear examples of strong artificial intelligence, the field of AI is rapidly innovating. Another AI theory has emerged, known as artificial superintelligence (ASI), super intelligence, or Super AI. This type of AI surpasses strong AI in human intelligence and ability. However, Super AI is still purely speculative as we have yet to achieve examples of Strong AI.
With that said, there are fields where AI is playing a more important role, such as:
The terms artificial intelligence, machine learning and deep learning are often used in the wrong context. These terms are frequently used in describing Strong AI, and so it’s worth defining each term briefly:
Artificial intelligence defined by John McCarthy (link resides outside IBM), 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."
Machine learning is a sub-field of artificial intelligence. Classical (non-deep) machine learning models require more human intervention to segment data into categories (i.e. through feature learning).
Deep learning is also a sub-field of machine learning, which attempts to imitate the interconnectedness of the human brain using neural networks. Its artificial neural networks are made up layers of models, which identify patterns within a given dataset. They leverage a high volume of training data to learn accurately, which subsequently demands more powerful hardware, such as GPUs or TPUs. Deep learning algorithms are most strongly associated with human-level AI.
To read more about the nuanced differences between these technologies, read “AI vs. Machine Learning vs. Deep Learning vs. Neural Networks: What’s the Difference?”
Deep learning can handle complex problems well, and as a result, it is utilized in many innovative and emerging technologies today. Deep learning algorithms have been applied in a variety of fields. Here are some examples:
Watson Assistant is the AI chatbot for business. This enterprise artificial intelligence technology enables users to build conversational AI solutions.
IBM Watson Assistant provides customers with fast, consistent and accurate answers across any application, device or channel.