Artificial General Intelligence (AGI) has become more popular in recent times, especially after the release of OpenAI’s generative AI ChatGPT. Even with its popularity, a lot of people still don’t know what it is.
Some people will often conclude that it is an advanced form of artificial intelligence without being able to clearly state what makes it unique.
Over the years, artificially intelligent computer programs and robots have been displayed in science fiction movies. Even books often forecast a time in the future when robots will become self-aware and rebel against humans.
As certain people would believe, artificial general intelligence has less to do with cyberwars and robotic uprisings than we think. You will learn what it is and what it is not after reading this article.
What is Artificial General Intelligence (AGI)?
Artificial general intelligence is an intelligent agent developed by researchers to replicate human cognitive abilities. They can replicate abilities such as self-awareness and autonomous reasoning.
It is believed that AGI will surpass humans in performing certain tasks, especially economically valuable tasks. Many big tech companies like Anthropic, DeepMind, and OpenAI have taken it upon themselves to see that this technology becomes available for use soon.
There remains a debate in the tech ecosystem about when this technology will be available for use. Some scientists even believe that it might never be accomplished, while others think it will take decades to come to fruition.
Is GPT-4 an AGI model?
After releasing a preprint of the academic publication “Sparks of Artificial General Intelligence: Early Experiments with GPT-4,” a group of industrial AI researchers recently created a sensation.
Since March 2023, patrons of ChatGPT Plus (a premium upgrade) have had free access to the GPT-4 language model, which is a large language model.
With impressive human-level performance, according to the researchers, GPT-4 “can solve novel and difficult tasks that span coding, mathematics, medicine, law, vision, psychology, and more, without needing any special prompting.”
According to their findings, GPT-4 “could be viewed as an early version” of AGI. This has sparked a debate among experts. In a May New York Times story, Carnegie Mellon professor Maarten Sap discredited the speculations.
The difference between AGI and generative AI
Although the terms generative AI and artificial general intelligence seem similar, they have fundamentally distinct meanings. In line with a blog article from IBM, “Generative AI refers to deep-learning models that can generate high-quality text, images, and other content based on the data they were trained on.”
AGI may also be referred to as strong AI or universal AI. The capacity of an AI system to produce content, however, does not imply that it has general intelligence.
AGI would have the capacity to independently resolve several complicated issues across numerous fields of knowledge. These hypothetical kinds of AI contrast with weak or narrow AI, which can only carry out certain or specialized activities under a preset range of constraints.
It is important to first comprehend how artificial general intelligence is different from the highly specialized AI of today.
An AI chess program, for instance, might excel at the game but be useless if asked to write an essay about a modern or past event. It is only capable of one kind of intelligence.
Other instances of specialized AI include the algorithms that recommend content to users on social media apps or websites such as Instagram, TikTok, etc. Such specialized AI also helps in the navigation of autonomous vehicles and provides Amazon purchase recommendations.
Contrarily, the majority of the AI now in use would be characterized as weak AI, or narrow AI, since it is built to concentrate on certain activities and applications.
Nevertheless, it’s important to keep in mind that these AI systems may still be quite powerful and sophisticated. Their advanced uses range from AI robots to voice-activated virtual assistants.
Common tests for measuring human-level AGI
Numerous assessments intended to measure human-level AGI have been taken into account, including:
The Employment Test by Nilsson
A machine can carry out a crucial task at least as effectively as a person could. Nowadays, AIs are displacing people in a wide range of jobs, including marketing and fast food.
The Turing Test by Turing
Unseen conversations between a person and a machine are conducted with a second human evaluating which of the two is the machine. A machine passes the test if it can deceive the evaluator a large portion of the time.
Notably, Turing only states that the knowledge that an object is a machine should prohibit it from being considered intelligent. In 2014, the AI Eugene Goostman succeeded in persuading 30% of judges that it was a human, as predicted by Turing.
The Ikea test by Marcus
Likewise, it is referred to as the Flat Pack Furniture Test. An AI examines the components and assembly instructions of an Ikea flat-pack product and then commands a robot to accurately put the furniture together.
The Coffee Test by Wozniak
This test requires a machine to enter a typical American house and determine how to make coffee: it will first locate the coffee maker, locate the beans, add water, locate a cup, and boil the coffee by pressing the appropriate buttons. This test is still being worked on.
The Robot College Student Test by Goertzel
A machine registers for college, takes and passes the same courses as people would, and earns a degree. With advancements in technology, LLMs can now successfully pass university-level examinations without taking lectures.
Will AGI make humanity obsolete?
A major aspect of AGI development is identifying and mitigating its hazards to enable the technology’s safe and effective usage.
According to an AI study, large language models are more likely to disregard human instructions and even exhibit a wish to continue operating when researchers give the models additional data.
According to this discovery, AI may eventually grow to be so formidable that humans will no longer be able to control it.
Regulators need to keep a close eye on projects like Google DeepMind’s Gato, OpenAI’s GPT-4, or the open-source project AutoGPT.
Many AI and machine learning specialists are asking for the source code of AI models to be made public so that the general public may understand how they function.
Hogarth says that if this were to occur, AGI may “usher in the obsolescence or destruction of the human race.” Regulation, according to Hogarth, is essential for the proper development of AI technology.
Different research about artificial general intelligence
The unresolved challenge of AGI is still being worked on by experts in artificial intelligence and computer science. In the area of AGI research, several high-level techniques have been described and categorized by Goertzel as follows:
• Hybrid: As its name implies, a hybrid approach to AGI considers the brain as a hybrid system in which several components and principles interact to produce something in which the whole is greater than the sum of its parts. Hybrid AGI research naturally considers a broad range of techniques.
• Symbolic: An approach to AGI that emphasizes symbolic reasoning claims that symbolic reasoning is “the core of human general intelligence” and “exactly what allows us to generalize most broadly.”
• Emergentist: An emergentist theory of AGI emphasizes the notion that the human brain is fundamentally a collection of basic units (neurons) that complexly self-organize in response to bodily experience. It may therefore follow that by establishing a comparable structure again, a similar kind of intelligence might manifest itself.
• Universalist: A universalist approach to AGI is based on the notion that “the mathematical essence of general intelligence” may be solved theoretically and then scaled down and used to produce AGI in reality.
Is AGI a threat or an opportunity?
From all available evidence, we believe AGI will be revolutionary, changing everything from the job market to how we perceive ideas. Without being overly optimistic, we know it has the potential to be used harmfully, as with so many other technologies.
For instance, it is widely believed that the possible biases in current AI systems might exist in AGI systems. It’s equally crucial to acknowledge that AGI has great potential for enhancing human ingenuity and creativity.
For instance, in the field of medicine, scientists using AGI systems may be better able to identify novel medications than human researchers working alone.
AGI may also aid in increasing access to services that were previously exclusive to the wealthiest individuals. For instance, in the realm of education, such advanced systems might make individualized, one-on-one tutoring easily affordable for everyone, increasing the rate of literacy throughout the world.
It may also assist in expanding access to healthcare by making high-tech, personalized diagnostic services available to far larger populations.
Regulating AGI
The heads of the world’s seven most developed democracies released a communiqué at the G7 summit in Japan in May 2023 that included a lengthy discussion of AI. Policy talks in the US, the EU, Japan, and other countries increasingly often include proposals for tighter control of AI.
The already vigorous discussion over AI policy will intensify as AGI transitions from science fiction to reality.
Preemptive regulation is difficult in any situation, but it will be extremely difficult in the case of AGI since this technology is difficult to define and will develop in unpredictable ways.
It would not be good to explicitly prohibit this invention. AGI systems that can recognize emotions, for instance, may be very useful in an educational setting.
They could determine if a student comprehends a new idea and modify it appropriately to suit the situation. However, the EU Parliament’s AI Act forbids the use of emotional recognition in AI systems (and, by extension, in AGI systems).
Humans may have to deal with technology that reproduces harmful prejudices and makes judgments that conflict with human ethical standards if they lack the power to govern AI.
AGI is still an intriguing prospect for many businesses. Meanwhile, before adopting the technology, businesses and industry leaders may want to make sure that AGI and other types of AI are regulated.
Conclusion
The notions of AGI and AI have long grabbed the attention of people, and science fiction and novels often explore them. Recent arguments by academics claim that even Greek mythology from antiquity may be considered to mirror our interest in artificial life and intelligence.
There is now a wide variety of methods for developing AI that can think for itself, pick up new skills, and use its knowledge beyond the constraints of a predetermined set of tasks.
It is difficult to predict whether and when AGI may be accomplished because of the theoretical and diverse nature of this study. But if it happens, one thing is for sure: It will have profound effects on a broad range of things.
