~2050

Artificial General Intelligence (AGI)

The definition and metric that determines whether computers have achieved human intelligence is controversial among the AI community. Gone is the reliance on the Turing test—programs can pass the test today, and they are clearly not intelligent.

So how can we determine the presence of true intelligence? Some measure it against the ability to perform complex intellectual tasks, such as carrying out surgery or writing a best-selling novel. These tasks require an extraordinary command of natural language and, in some cases, manual dexterity. But none of these tasks require that computers be sentient or have sapience—the capacity to experience wisdom. Put another way, would human intelligence be met only if a computer could perform a task such as carrying out a conversation with a distraught individual and communicating warmth, empathy, and loving behavior—and then in turn receive feedback from the individual that stimulates those feelings within the computer as well? Is it necessary to experience emotions, rather than simulate the experience of emotions? There is no correct answer to this, nor is there a fixed definition of what constitutes “intelligence.”

The year chosen for this entry is based upon broad consensus among experts that, by 2050, many complex human tasks that do not require cognition and self-awareness in the traditional biochemical sense will have been achieved by AI. Artificial general intelligence (AGI) comes next. AGI is the term often ascribed to the state in which computers can reason and solve problems like humans do, adapting and reflecting upon decisions and potential decisions in navigating the world—kind of like how humans rely on common sense and intuition. “Narrow AI,” or “weak AI,” which we have today, is understood as computers meeting or exceeding human performance in speed, scale, and optimization in specific tasks, such as high-volume investing, traffic coordination, diagnosing disease, and playing chess, but without the cognition and emotional intelligence.

The year 2050 is based upon the expected realization of certain advances in hardware and software capacity necessary to perform computationally intense tasks as the measure of AGI. Limitations in progress thus far are also a result of limited knowledge about how the human brain functions, where thought comes from, and the role that the physical body and chemical feedback loops play in the output of what the human brain can do.

SEE ALSO The “Mechanical Turk” (1770), The Turing Test (1951)

Artificial general intelligence refers to the ability of computers to reason and solve problems like humans do, in a way that’s similar to how humans rely on common sense and intuition.