AI is largely responsible for the world’s rapid pace of change. Since AI has become so pervasive in our professional and personal lives, more innovative and automated solutions are now a necessity. Researchers and scientists are working tirelessly to eliminate all technological limitations and develop new branches of AI to automate processes. Artificial general intelligence (AGI) is one such cutting-edge branch of AI.
AI has been split into two separate groups:
- artificial narrow intelligence, which is what we have now, and
- artificial general intelligence, which is what we want to get.
What is artificial general intelligence?
A machine with AGI would be capable of comprehending the world as well as a human and possess the same capacity for learning how to perform a wide variety of tasks. AGI’s primary functions are to create fully capable artificial systems, train them to experience, adapt to new situations, and perform human-like tasks efficiently. Additionally, AGI is more accurate than any other technology at simulating human tasks. When given specific tasks in advance, these systems can even outperform humans in terms of efficiency. Simply put, an AGI is a machine that possesses general intelligence comparable to that of a human being and is capable of solving any problem.
What could an AGI do?
In theory, an AGI could perform any task that a human could, as well as many more that a human could not. At the very least, an AGI would be able to combine human-like, flexible thinking and reasoning with computational advantages. This intelligence could be used to control robots and mobile devices as if they were people, resulting in a new breed of machines. These artificial intelligences would eventually be able to take over every role currently performed by humans. However, the arrival of AGI will almost certainly render human labour obsolete.
There has been a gradual rise in the number of people who are interested in AGI in the wider AI community.
- Springer published “Artificial General Intelligence” as an edited volume in 2005. (Goertzel and Pennachin, 2005).
- In 2006, Bethesda, Maryland hosted the first formal research workshop on “AGI” (Goertzel, 2013).
- AGI has been described by Joscha Bach as a quest to create “synthetic intelligence” (2009, Bach).
- Since its inception, the AGI community has included researchers working in a variety of different directions, some developing cognitive architectures inspired by cognitive psychology and neurobiology.
- Certain AGI cognitive architecture research builds on concepts from classic AI cognitive architectures such as SOAR (Laird, 2012) and GPS (Newell et al., 1959).
- Ray Solomonoff (1964) and other early thinkers of formal intelligence theory are a big influence on modern AGI’s math foundations.
In the years since, a wide range of researchers have come together to work on AGI.
What features would enhance AI to AGI?
Perception sensory:
While deep learning has enabled significant advances in computer vision, AI systems are still a long way from achieving human-like sensory perception. For example, deep learning-trained systems still exhibit inconsistent colour perception: self-driving car systems have been fooled by small pieces of black tape or stickers on a red stop sign. AI systems are not yet capable of replicating this uniquely human perception.
Manual dexterity:
Any human being is capable of retrieving a set of keys from his or her pocket. Very few of us would delegate that task to any of the robot manipulators or humanoid hands we see. This issue is being addressed by researchers in the field. Recent research demonstrated how reinforcement learning can be used to teach a robot hand to solve a Rubik’s cube.
Natural language comprehension:
Humans communicate their abilities and knowledge via books, articles, blog posts, and, more recently, how-to videos. For AI to be fully comprehended and able to consume these sources of information, Without this foundation of common-sense knowledge, AI will be unable to function in the real world.
Solving problems:
In any general-purpose application, a robot must be capable of diagnosing and resolving problems. A home robot would need to recognise when a light bulb is blown and take the appropriate action, such as replacing the bulb or contacting a repair person. To be able to do these things, the robot must have some of the common sense we talked about before, or it must be able to run simulations to figure out what is possible, plausible, and likely.
Navigation:
When combined with capabilities such as simultaneous localization and mapping (SLAM), GPS has made significant advancements in this field. Years of work are still required to develop robust systems capable of functioning without human priming. Academic demonstrations to date have fallen far short of this objective.
Creativity:
While machines have shown that they can draw pictures and write music, they still need to be improved to be as creative as humans are.
Social and emotional engagement:
To be successful in our world, humans must desire to interact with robots and artificial intelligence, rather than fear them. The robot must be able to read people’s facial expressions and tone changes to figure out what their real feelings are.
Milestones in AGI:
In the 1950s and 1960s, the pioneers of AI were primarily concerned with developing hardware or software capable of simulating human-like general intelligence. Since then, the field has shifted its emphasis to pursuing discrete capabilities or specific practical tasks.
- As far back as 1997, Mark Gubrud used the term “artificial general intelligence” to talk about how fully automated military production and operations might work.
- In 2002, Shane Legg and Ben Goertzel reintroduced and popularised the term.
- Pei Wang and Ben Goertzel defined AGI research activity in 2006 as “producing publications and preliminary results.”
- In 2009, the Xiamen University’s Artificial Brain Laboratory and OpenCog jointly organised the first summer school on AGI in Xiamen, China.
- Todor Arnaudov taught the first university courses in 2010 and 2011 at Plovdiv University in Bulgaria.
- In 2018, MIT hosted an AGI course organised by Lex Fridman and featuring a variety of guest lecturers.
However, a small number of computer scientists are engaged in AGI research, and many of them contribute to an annual series of AGI conferences. The research is extremely varied and frequently innovative in nature.
Conclusion
Currently, the majority of AI examples that you hear about—from chess-playing computers to self-driving cars—rely on deep learning and natural language processing. However, existing AI systems can only do the tasks they’ve been told to do perfectly, but they can’t do any other task with the same level of precision. AGI is a primary objective of some AI research and a frequently discussed subject in science fiction and futurology. AGI is also referred to as strong AI, complete AI, or intelligent action in general. Without a doubt, the future of AI is an exciting career.
Source: indiaai.gov.in
