Artificial general intelligence (AGI) describes AI that is capable of learning an intellectual task in the same way humans do.
Understanding artificial general intelligence
Artificial general intelligence is the representation of general human cognitive abilities in software that enables AI to solve problems in the same way a person does.
The precise definition of AGI varies from one expert to the next since each approaches the subject of human intelligence from a different perspective.
Psychologists, for example, may define it as the ability to adapt and survive while computer scientists may consider intelligence to be more related to goal achievement.
In any case, AGI is believed to be a form of strong artificial intelligence. This form contrasts with weak or narrow AI that is used to perform specific tasks or solve specific problems.
Autonomous vehicle technology and IBM’s Watson supercomputer are two such examples.
At present, AGI is a theoretical construct and remains the stuff of science fiction. When Gato was released by Alphabet subsidiary Deepmind in May 2022, it was touted as a “generalist agent” that could perform over 600 tasks from captioning an image to driving a robot.
Gato is probably the most advanced AI system in the world, but it can only make inferences from the information stored in its giant database.
For AGI to come to fruition, many argue it will require innovators to do more than simply force algorithms to parse more data.
On the subject of when artificial general intelligence may materialize there is much debate. Some academics believe AGI is decades away, while others predict the technology will not be developed this century.
Some, such as MIT roboticist Rodney Brooks, argue that AGI will not arrive until after the year 2300.
Characteristics of artificial general intelligence
While artificial general intelligence remains theoretical, there is scope that the performance of an AGI system will not only be indistinguishable from a human but far exceed it.
This is because these systems will likely possess comprehensive cognitive computing capabilities and the ability to process vast data sets at incredible speeds.
Nevertheless, some of the human characteristics artificial general intelligence must be able to replicate include:
- Common sense.
- The ability to understand cause and effect.
- Transfer learning – the application of knowledge learned from completing one task to solving a different but related problem.
- Abstract thinking.
- Sensory perception – this includes subjective color perception and depth perception in static images.
- Fine motor skills.
- Superior navigation skills – while existing GPS can pinpoint a specific location, it is envisioned that AGI will be able to better project movement through physical spaces.
- Natural language understanding (NLU) – this would require AGI to possess a level of intuition enabling it to understand human language which is heavily context-dependent.
- Other capabilities such as the comprehension of belief systems, symbols, and metacognition which includes self-awareness and critical thinking.
- Artificial general intelligence (AGI) describes AI that is capable of learning an intellectual task in the same way as humans do.
- While recent advancements in AI technology such as Gato have been commendable, AGI is at present a theoretical construct. Some experts believe it will take centuries for artificial general intelligence to be developed.
- With access to large datasets and superior processing power, there is scope that AGI may be able to outperform humans in the future. Before that happens, however, scientists must be able to replicate difficult human characteristics such as abstract thinking, fine motor skills, and sensory perception, among many others.