What we know so far about OpenAI's artificial general intelligence

ChatGPT's parent company aims to have its first AI agents in the workforce this year

AGI is an advanced form of artificial intelligence capable of performing any intellectual task that a human can do. Reuters

OpenAI chief executive Sam Altman has announced that the company is confident in its ability to develop artificial general intelligence, with the first AI agents expected to enter the workforce this year.

These agents could boost productivity and transform industries by performing tasks traditionally handled by humans. AI agents are an intermediate step between current AI systems and the future AGIs.

“AGI could be the most impactful technology in human history,” said Mr Altman, who co-founded OpenAI in 2015.

The announcement came as the global race to build AGI intensifies, with tech giants investing billions of dollars to achieve breakthroughs in AI. Mr Altman emphasised the importance of gradually deploying such tools to ensure societal adaptation and equitable outcomes.

However, the move raises questions about which industries will see the earliest adoption of AI agents, how they will co-exist with human workers, and whether OpenAI can maintain its leadership in this competitive field.

The National dives into the nitty gritty of AGI, while considering its future in 2025.

What is AGI?

AGI is an advanced form of artificial intelligence capable of performing any intellectual task that a human can do. Unlike current AI systems, which are designed for narrow, specific purposes like language translation or image recognition, AGI aims to understand, learn and adapt to a range of jobs across various industries.

Mr Altman defines AGI as a technology that can reason, learn and operate across all areas of human cognition. For instance, an AGI system could write software, design complex architecture, and solve quantum computing equations without requiring separate programming for each task.

Sam Altman, chief executive of OpenAI. EPA

“We love our current products, but we are here for the glorious future. With superintelligence [AGI], we can do anything else,” Mr Altman said.

“Superintelligent tools could massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own, and in turn massively increase abundance and prosperity.”

Difference between AI and AGI

Current AI systems, often called narrow AI, are designed to perform specific tasks. For example, Google’s generative AI tool Gemini or OpenAI’s ChatGPT focuses on generating text, images and codes but they lack the ability to generalise across different tasks.

Whereas AGI is not limited to specific tasks but can apply knowledge and reasoning to new, unfamiliar problems without the need for human intervention. While narrow AI might plan an adventure trip for a user, AGI could personalise it further, tailoring the itinerary to the user’s preferences, past choices or specific audience needs.

Microsoft-backed OpenAI defines AGI as “AI systems that are generally smarter than humans”.

“AGI has the potential to give everyone incredible new capabilities … we can imagine a world where all of us have access to help with almost any cognitive task, providing a great force multiplier for human ingenuity and creativity,” it added.

Will OpenAI be the first to build AGI?

OpenAI is trying to position itself as a leader in the race to achieve AGI but it is not alone in this competition.

Companies like Alphabet-owned AI research lab DeepMind, Anthropic, the company founded in 2021 by former OpenAI employees, Elon Musk’s Grok and global tech giants such as Apple, Nvidia, Google and Microsoft are investing heavily in similar research.

Last month, Google launched Gemini 2.0, its most advanced generative AI model yet, marking a shift from information retrieval to action-oriented argentic AI – systems that can think, plan, and take action on users' behalf.

Argentic AIs are systems that can take initiative, make decisions, and perform actions on behalf of users, guided by human input and supervision. An argentic AI could automatically book a hotel, suggest activities, make dinner reservations, and provide a schedule – all based on users’ past preferences.

The number of ChatGPT's weekly active users has surpassed 300 million. AFP

OpenAI’s past progress with models like GPT-4 and a user base of more than 300 million weekly active users gives it an industry edge, with Mr Altman emphasising that the company now “knows how to build AGI”.

Will AGI be a reality in 2025?

The AGI market size is expected to reach $52 billion by 2032, growing at a compound annual growth rate of 37.5 per cent between 2024 and 2032, according to SNS Insider research. It stood at $3 billion in 2023.

Advancements in machine learning, deep learning, and robust computing infrastructure are fuelling the global push for AGI solutions.

Mr Altman’s latest claim that AI agents could join the workforce by 2025 hints at AGI-like systems becoming functional soon but whether this will actually happen remains debatable.

“We are now confident we know how to build AGI as we have traditionally understood it. We believe that, in 2025, we may see the first AI agents join the workforce,” Mr Altman said.

What experts say?

Some industry experts consider AGI as a logical next step of generative AI in the wake of recent advancements in technology, while others argue the timeline for practical AGI remains overly optimistic.

AI researcher Gary Marcus has expressed concerns about the rapid achievement of human-level reasoning and adaptability in AI systems. He noted that while AI has made significant strides, current models lack the deep understanding and reasoning inherent to human cognition. He suggested that achieving humanlike intelligence in AI is still a work in progress.

While DeepMind's Demis Hassabis believed that advanced AI systems could revolutionise industries sooner than expected.

Stefan Leichenauer, vice president engineering at SandboxAQ said so far generative AI has been dominated by LLMs that use language as a fundamental concept to solve problems, and along with that come a number of challenges that are tough to avoid.

“A purely language-based model is subject to hallucinations and cultural bias … the AI that will be used increasingly for breakthroughs in science and technology, Quantitative AI, has a different flavour.”

“It will be based on fundamental principles, such as verifiable mathematical equations or real experimental data, that are unbiased. In areas such as life science, healthcare, or next-generation materials, this approach is critical,” Mr Leichenauer said.

Why OpenAI’s $200-a-month Pro subscription struggling?

As the company moves towards achieving its AGI ambitions, Mr Altman announced on Monday that OpenAI’s top-tier $200-per-month Pro subscription is currently not generating profits.

“Insane thing: we are currently losing money on Open AI pro subscriptions! people use it much more than we expected,” he wrote on X.

He also added that he personally decided the price and thought the company would make some money out of this initiative. However, he did not provide the numbers of Pro subscribers and logic behind choosing this price point.

One of the potential reasons could be high operational costs as running these models usually requires a huge amount of computing power, cloud infrastructure, storage and energy. If the number of users exceeds OpenAI’s expectations, operational costs might top revenue from subscriptions.

Announced last month, ChatGPT Pro targets researchers, engineers, and other individuals who use research-grade intelligence to accelerate their productivity.