What Is Artificial General Intelligence?

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 that it wasn’t initially trained to do.

If artificial general intelligence 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 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.”

 

What Is Artificial General Intelligence?

Artificial general intelligence (AGI), also known as strong AI, is a hypothetical type of AI that can understand, learn and apply knowledge on a level that matches or surpasses that of humans. 

Unlike the “narrow” or task-specific AI we have today, AGI is envisioned to go beyond these capabilities to perform “a broad array of cognitive tasks with flexibility and generalization comparable to human reasoning,” as stated in a 2025 study.

The goal of AGI is to create machines that not only mimic human behavior in certain contexts, but also replicate the general intelligence that enables this behavior in the first place.

Although AGI remains a theoretical concept today, it stands as a long-term goal — and arguably the next big milestone — of AI research and innovation.

 

Artificial General Intelligence vs. Artificial Intelligence

Artificial general intelligence is a more advanced form of artificial intelligence.

• 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 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.

 

Types of Artificial Intelligence: Narrow AI vs. AGI vs. ASI

Artificial general intelligence is considered one of the three main types of AI. 

1. Weak AI (Narrow AI)

Weak AI, also known as narrow AI or artificial narrow intelligence (ANI), can only perform 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 store past information to improve their performance over time. 

2. Strong AI (Artificial General Intelligence)

AGI, or strong AI, can replicate human intelligence. When strong AI learns how to complete one task, it can take this knowledge and apply it to other tasks. Strong AI can then take on challenges it was never trained for, demonstrating the kind of advanced problem-solving and adaptability associated with humans.  

3. Artificial Superintelligence (ASI)

Artificial superintelligence (ASI) is a theoretical form of AI that would be able to learn at a rapid rate to the point where it surpasses the abilities of humans. ASI is seen as the technology necessary to develop self-aware AI. In this state, AI would be able to act according to its own will and disregard instructions or its intended purpose. 

While AI tools today mostly belong to the weak AI category, some believe we are inching closer toward achieving artificial general intelligence.  

 

Key Characteristics of Artificial General Intelligence

Artificial general intelligence is marked by its ability to perform a wide range of tasks with human-like reasoning and understanding.

Here’s five key characteristics of AGI that would make it distinct from today’s AI technologies:

1. Cross-Domain Generalization

AGI can transfer knowledge and skills learned in one area to entirely different and unfamiliar tasks. This ability to generalize across domains allows AGI to solve problems creatively and flexibly, much like a human would in new situations.

2. Autonomous Learning

AGI can learn autonomously from raw data and experiences, without needing labeled training sets or constant human intervention. It would adapt over time by observing its environment, making inferences and improving in tasks based on learned successes and failures.

3. Logical Reasoning and Problem-Solving

AGI is capable of logical reasoning, problem-solving and decision-making on a similar level as humans. This allows it to analyze new problems, consider alternatives and develop unique solutions to scenarios, even when given incomplete or ambiguous information.

4. Natural Language Understanding

AGI has a deep understanding of human language and its nuances, including aspects like human tone and intent. It can engage in meaningful and fluid conversations, comprehend complex instructions and respond appropriately to various social and cultural contexts.

5. Adaptability and Goal-Directed Behavior

AGI’s goal-oriented behavior allows it to adjust to changing environments in real time and pursue goals independently to achieve desired results. This means it could determine its own priorities, balance competing actions and make decisions aligned with long-term goals, rather than only executing preset instructions.

 

What Could Artificial General Intelligence Do?

Current forms of AI are able to master a specific task they’re programmed to undertake. AGI, in theory, takes this capability a step further, adapting to unfamiliar situations it was never trained to handle. This opens the door for many more applications:  

• Healthcare: AGI could analyze massive volumes of patient data to identify at-risk patients, predict future diseases and design personalized treatments.
• Education: AGI could curate a unique curriculum for students based on their individual academic performance and learning style.
• Customer service: AGI could use past calls and demographic info to tailor service to each customer, anticipate questions and take proactive measures before issues occur.   
• Finance: AGI could compile information to enhance the accuracy of financial models, predict market behavior and execute informed trades based on real-time insights.
• Self-driving cars: AGI could collect real-time information (on weather, traffic patterns, etc.) from sensors and make instant adjustments to adapt to various scenarios.  
• Programming: AGI could understand coding logic to not only generate code, but also make recommendations and design entire functions to fulfill particular needs.  
• Manufacturing: AGI could process large amounts of data gathered from sensors to predict machine issues and alert teams before equipment breaks down.

 

Examples of Artificial General Intelligence

While artificial general intelligence doesn’t exist, the following examples show that AGI may not be that far off. 

ChatGPT-4o 

GPT-4o builds on the foundation laid by previous GPT models — namely GPT-4 — with the goal of realizing “more natural human-computer interaction.” The model can process text, visuals and audio and respond via text or its own voice to sustain in human-like conversations. 

“Given the breadth and depth of GPT-4’s capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system,” according to a 2023 study.

Other models developed at top companies like Google, Anthropic and xAI also appear to be approaching AGI. But, of course, without an industry-wide definition or benchmark, it’s unclear how close any of them truly are — or how far they still have to go.

Self-Driving Cars

Many cars include autonomous features, like Honda’s traffic jam feature that can adapt to crowded conditions on the road. Meanwhile, Waymo’s autonomous ride-hailing service demonstrates complete autonomy while transporting customers. 

AlphaFold 3

Designed by Google DeepMind, AlphaFold 3 doesn’t just predict protein structure but can also predict the structures of life’s building blocks, including DNA and RNA. This makes it possible to generate models of various molecular structures and accelerate drug development.

AI Music Generators

AI music generators are beginning to compete with human musicians, producing songs that generate plenty of buzz online. For example, Suno can supplement AI-generated lyrics with vocals and instrumentals, crafting all the elements of a song on its own.

RelatedCan AI Make Art More Human?

 

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 its arrival will transform daily life in countless ways: 

• Increased productivity: AGI could adapt to new tasks and roles, saving teams from having to repeatedly train it and freeing up workers to handle other challenges.  
• Enhanced problem-solving: AGI could use its advanced computing power to address global issues like climate change, widespread disease and supply-chain needs. 
• Faster healthcare: AGI could quickly diagnose diseases and design new treatments, resulting in faster and more personalized healthcare. 
• In-depth interaction: AGI could offer sharp insights and sound advice, becoming a more active collaborator and even taking on roles like mentor.  
• Original creativity: AGI could use flexible thinking to craft its own movie scripts, songs, videos and other creative works with little to no prompting. 

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,” Amruth Laxman, co-founder of 4Voice, told Built In. “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, and AGI could come with even more consequences:  

• Ethical questions: There’s reason to question whether AI can understand human ethics, so humans may have to confront an AI that doesn’t follow human ethical standards.
• Social inequities: AGI requires the kind of capital and resources only large corporations have, concentrating even more power in the hands of a few businesses.  
• Lack of legal safeguards: Laws still lag behind weak AI, so AGI would likely have no legal limits and could potentially be used for malicious purposes.    
• Job losses: The development of an AI that can mimic and surpass human abilities may trigger fears of job losses due to automation.

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, chief technology officer 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.”

RelatedAre You Sure You Can Trust That AI?

 

Who’s Leading the Race to AGI?

With the AI industry rapidly growing, several tech giants and AI-focused startups are currently competing to be the first to develop advanced AI solutions — including artificial general intelligence.

Below are some of the biggest players leading the race to achieve AGI:

OpenAI

OpenAI, developer of ChatGPT, is on a mission to ensure artificial general intelligence systems “benefit all of humanity.” The company has famously taken the spotlight with its GPT foundation models and introduction of agentic AI systems (specifically ChatGPT agent), which has been noted by OpenAI CEO Sam Altman as a real-feeling step toward AGI.

Google

Google DeepMind, an AI research lab and subsidiary of Alphabet Inc., continues to push foundational AI research with systems like AlphaGo, AlphaFold and AlphaGeometry. In 2025, Google DeepMind announced it is “exploring the frontiers of AGI,” with CEO Demis Hassabis stating that AGI is likely to arrive around 2030.

Anthropic

Anthropic, developer of Claude, focuses on building reliable and interpretable AI systems. Its Claude 4 models set modern standards for advanced reasoning, coding and AI agents, with Claude Opus 4 in particular demonstrating sustained performance on complex, long-running tasks and agent workflows.

xAI

Elon Musk’s xAI aims for AGI through its Grok models and their deep computation capabilities. The Grok 4 model “represents a leap in frontier intelligence,” showing capabilities in complex reasoning through scaled reinforcement learning and native tool use. xAI claims Grok 4 outperforms all other models on the ARC-AGI benchmark, which measures AI progress toward AGI.

Meta

Meta — known for its Llama models — announced the foundation of Meta Superintelligence Labs (MSL) in 2025, a company division that will build toward AGI or “personal superintelligence” (as it’s referred to by Meta). In the MSL announcement, Meta CEO Mark Zuckerberg stated that Meta would begin research on its next-generation AI models to “get to the frontier in the next year or so.”

 

History of Artificial General Intelligence

The pursuit of AGI has evolved alongside the broader history of artificial intelligence. While the concept has been around for decades, real progress has only emerged in the 21st century thanks to advancements in machine learning, computing power and data availability.

Here are some of the key milestones on the path toward AGI we’ve achieved to date:

2000s: Beginnings of Modern AI

• 2006: Computer scientist and “Godfather of AI” Geoffrey Hinton and other researchers revive interest in deep learning, a type of AI that enables AI systems to learn from large amounts of data.
• 2011: IBM Watson defeats human champions on Jeopardy!, demonstrating advancements in language understanding.

2012 to 2018: Deep Learning Boom

• 2012: An ImageNet breakthrough, known as AlexNet, propels neural networks into mainstream AI research, highlighting the power of large-scale data and compute power.
• 2014 to 2018: The development of models like Google DeepMind’s AlphaGo and Google’s Transformer architecture lays the groundwork for more generalizable learning systems.

2018 to 2022: Rise of Foundation Models

• 2018: OpenAI releases GPT-1 model, marking the start of the large language model (LLM) era.
• 2019 to 2020: OpenAI releases GPT-2 and GPT-3 models, demonstrating that scaling models and training data can lead to general capabilities, sparking discussion about AGI’s possibility.
• 2022: Google DeepMind’s Gato blurs the lines between narrow and general AI by showing cross-task versatility.

2023 to Present: The AGI Race

• 2023: OpenAI releases GPT-4, Elon Musk founds xAI to compete directly with OpenAI in the AI race.
• 2024: Anthropic’s Claude 3 and Google’s Gemini 1.5 models signal significant strides in AI reasoning and memory. 
• 2025: Companies like OpenAI, Google, Anthropic, xAI and Meta openly declare ambitions for achieving AGI, with researchers forecasting AGI timelines within the coming decades. DeepSeek releases DeepSeek-R1 model, surpassing some capabilities of GPT-4 and other foundation models. OpenAI releases ChatGPT agent, which can carry out tasks using its own virtual computer.

 

Future of Artificial General Intelligence

According to a TIME article, some forecasters predict AGI could exist as early as 2030, while many others don’t foresee AGI being achieved until decades later at the earliest. But forms of advanced AI continue to bring the field closer to AGI, with Google DeepMind’s AlphaGeometry 2 being seen as an AGI milestone due to its performance on Olympiad math questions and OpenAI claiming it is close to building AI that can reason.   

If AGI is ever realized, it 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. Complex tasks and workflows would become AI-powered, saving organizations time and money. More ambitious views of AGI even envision it helping humans address large-scale problems like climate change.

Of course, concerns remain about artificial general intelligence being developed without any laws or policies that could hold companies accountable. Researchers have responded by calling for “ethical frameworks and governance mechanisms” to keep the technology in check. Regulations for current AI technologies are also on the horizon, with the EU AI Act being rolled out in the coming years.     

Steps taken to monitor weak AI could open the door for more robust AI policies that can better prepare society for AGI and even more intelligent forms of AI. Governments and societies may then want to take proactive measures to ensure AI organizations prioritize the common good, so people can enjoy the benevolent aspects of self-aware AI and a higher quality of life.   

“Cognitive power is going to flow from sentience,” computer scientist and AI researcher Selmer Bringsjord previously told Built In. “And when power is available and not sufficiently controlled — self-controlled or controlled by others — really, really bad things can happen.”

Frequently Asked Questions