Artificial general intelligence (AGI) is a hypothetical form of artificial intelligence in which a machine can learn and think like a human. For this to be possible, AGI would need self-awareness and consciousness, so it could solve problems, adapt to its surroundings and perform a broader range of tasks.
“The best way to describe it is that AGI implements logic into the process rather than just applying an algorithm or coded process,” Amruth Laxman, co-founder of 4Voice, told Built In.
Artificial General Intelligence Definition
Artificial general intelligence is AI that can learn, think and act the way humans do. Although AGI has yet to be created, in theory it could perform a wider array of tasks than weak artificial intelligence and perform creative actions that previously only humans could.
If artificial general intelligence (sometimes called strong AI) sounds like sci-fi, that’s because it still is. Existing forms of AI haven’t quite reached the level of AGI — but developers are still working to make it a reality.
AGI vs. AI: What’s the Difference?
AGI is a subcategory of AI, and the former can be seen as an upgraded version of the latter.
Artificial intelligence is often trained on data to perform specific tasks or a range of tasks limited to a single context. Many forms of AI rely on algorithms or pre-programmed rules to guide their actions and learn how to operate in a certain environment.
Artificial general intelligence, on the other hand, is able to reason and adapt to new environments and different types of data. So instead of depending on predetermined rules to function, AGI embraces a problem-solving and learning approach — similar to humans. Because of its flexibility, AGI is capable of handling more tasks in different industries and sectors.
Weak AI vs. General AI
Weak or narrow AI is the type of artificial intelligence that powers self-driving vehicles, algorithmic image generators and chatbots. Simply put, it is AI that performs a limited range of tasks.
Two subsets of AI fall under the weak AI category: reactive machines and limited memory machines. Reactive machines can react to immediate stimuli, but cannot store or learn from memories of past actions.
Limited memory machines can, as the name suggests, store past information in order to improve its performance over time. They represent the majority of AI tools available today, although artificial general intelligence principles are slowly being incorporated into limited memory products.
Artificial General Intelligence Examples
AI that learns, thinks and behaves just like a person still doesn’t exist. However, weak AI tools incorporating some elements of artificial general intelligence, or that push the boundaries of limited memory, do exist.
“A couple of the best-known ways are self-driving cars and supercomputers like IBM’s Watson,” Laxman said. Creating music with AI programs and use in military supercomputers and quantum computers to strategize war and crisis scenarios are other applications, he added.
Some forms of weak artificial intelligence in use might seem advanced enough to reach artificial general intelligence status — think Sophia the robot. The field of generative AI, in particular, has made giant strides with tools like ChatGPT, suggesting that AGI isn’t far away.
“AGI doesn’t exist today in the way we think about it,” Wayne Chang, cofounder of Digits, told Built In. “However, the speed of innovation towards AGI is accelerating. In its ideal state, AGI would perform tasks that are identical to or surpass those that a human would perform.”
Benefits of Artificial General Intelligence
The development of AI technology is progressing in leaps and bounds. Artificial general intelligence might not be here today, but when it arrives, it will transform life as we know it.
In many ways, this transformation will mean huge benefits for society. Artificial general intelligence will be able to scan all preexisting information available in places like the internet to solve some of the world’s most pressing problems.
“AI has many positive uses now but has enormous future potential,” Laxman said. “It could, in the future, find a cure for chronic illnesses like cancer or resolve issues like overburdened utility infrastructure.”
Risks of Artificial General Intelligence
For all its potential benefits, artificial general intelligence doesn’t come without risks. Already, AI is challenging our perception of the world and what makes us human. The development of an AI that can mimic and surpass our own abilities may trigger fears of job losses due to automation and other potential dangers.
Though AI has machine precision, it isn’t immune to human error. Within weak AI, issues have already arisen where embedded systems have been built with biased data. This can result in AI making erroneous or, at worst, discriminatory decisions.
“Existing attempts at large AI models are trained with unfiltered and unreviewed data,” Chang said. “Because of this, a major concern is biased data, which can in turn compound within the systems and be exaggerated through the models.”
Artificial general intelligence does come with its dangers. But as long as the humans at the wheel have good intentions, Arnold Liwanag, CTO at AI company Tealbook, isn’t worried.
“AI is a tool,” Liwanag said. “The risk is only as great as the intent people have when using these tools.”
Future of Artificial General Intelligence
Although AGI is still a ways off, rapid progress in the AI field has brought technology closer to achieving AGI. Some predictions even have computers reaching the level of human intelligence as early as 2029.
The realization of AGI would mean AI that could act on abstract thinking, common sense, background knowledge, transfer learning and cause and effect. This would open up the possibilities for numerous industries. AGI could perform surgeries in the medical field and bring about autonomous cars in the automotive industry. More ambitious views of AGI even have the technology helping humans address large-scale problems like climate change.
With its far-reaching abilities, AGI could automate tasks that typically involve the kind of abstract thinking that only humans are capable of performing. Complex tasks and workflows would become AI-powered, saving organizations more time and money.
Ethical Considerations of Artificial General Intelligence
The benefits of AGI sound promising, but society needs to approach AGI with caution.
While AGI hints at a reality where AI can learn more difficult concepts unique to humans, there’s reason to question whether AI can understand human ethics. Even advanced technologies like ChatGPT have demonstrated harmful biases related to race and gender.
Current issues occurring before the existence of AGI may hint at larger problems emerging under artificial superintelligence — AI that transcends human intelligence. Without the ability to control AI, humans may have to face technology that reproduces devastating biases and makes decisions that go against human ethical standards.
Even if humans do control AGI, there’s the question of who gets to control it and what kind of power that gives them. For example, Meta already boasts a vast social media business, and the possibility of achieving AGI potentially puts more power in the hands of CEO Mark Zuckerberg. If Zuckerberg follows in the footsteps of AI competitors and decides not to open source Meta’s models, the ability to build AGI becomes further concentrated in the hands of a few corporations.
AGI remains an exciting proposition for many industries, but AI is ultimately a reflection of its creator’s intentions. Organizations and industry leaders may want to ensure AGI and other forms of AI are developed for benevolent purposes before embracing the technology.
Frequently Asked Questions
What is the difference between AGI and AI?
AGI is a subset of AI and is theoretically much more advanced than traditional AI. While AI relies on algorithms or pre-programmed rules to perform limited tasks within a specific context, AGI can solve problems on its own and learn to adapt to a range of contexts, similar to humans.
How close are we to AGI?
There’s no timeline for when AGI will be achieved. While companies like OpenAI and Meta are pursuing the development of AGI technologies, these remain a ways off.
Is AGI actually possible?
The possibility of AGI is debatable. Advancements have been made in the field of AI, but AGI remains purely theoretical at this point.