Artificial General Intelligence in 2025: How Close Are We?

The buzz around AI these days often revolves around artificial general intelligence (AGI) – the idea of a machine that matches human thinking across any task. In early 2025, even industry leaders have stoked excitement about AGI’s timeline. 

OpenAI CEO Sam Altman wrote that “we are now confident we know how to build AGI [as traditionally understood]” (OpenAI blog). And Google DeepMind’s Demis Hassabis told CBS News that AGI is “just five to 10 years away” (CBS News interview).

Yet many experts urge caution. “We can’t presume that we’re close to AGI because we really don’t understand current AI, which is a far cry from the dreamed-of AGI,” notes author and analyst HP Newquist (InformationWeek). In other words, today’s AI systems (like ChatGPT and Midjourney) are powerful but narrow – they excel at specific tasks but lack true understanding.

A large survey of over 2,700 AI researchers in 2025 estimated only about a 10% chance that AI would surpass human-level performance on most tasks by 2027 (AI Impacts 2023 expert survey). In this view, AGI by 2025 looks very unlikely. ZeroGPT CEO Edward Tian agrees: he told InformationWeek that “because of all these questions and our limited capabilities… I think that 2025 isn’t realistic”.

Why do experts remain skeptical? Current AI models can process text, images or other data, but they can’t learn or reason continuously in the way humans do. Columbia professor Max Li points out that existing models often give inconsistent or biased results, and “we do not even know why [some components] work, nor why they do not work”.

In short, today’s systems lack common sense and self-awareness. To truly achieve AGI, an AI would need to continuously learn from experience, form judgments about people and environments, and even reconsider what it has learned – abilities that remain science-fiction for now.

In the meantime, AI development is still accelerating. New tools like Google’s Gemini 2.0 and OpenAI’s upcoming GPT-5 promise ever more human-like interactions (for example, GPT-5 is rumored to unify text, voice and image “tools” into one model. These advances make AI feel smarter, but they are still narrow intelligence (ANI) – none can truly understand or imagine like a person.

Most experts say true AGI is still years if not decades away. As HP Newquist sums up: today’s AI is “a far cry from the dreamed-of AGI”, and closing that gap will take big breakthroughs. For 2025, that means AGI is more hope than fact. The race toward general intelligence continues, but for now AGI remains an open question rather than a sure thing.

---