Are We Close to Artificial General Intelligence? A Balanced Look at the Timeline

Dr. Sunday Oludare Ogunlana

Artificial General Intelligence (AGI) denotes a system that can match or exceed human capabilities across a wide array of tasks. The ascent of large language models and multimodal systems such as GPT‑4 and Gemini has fueled hopes that AGI might be just around the corner. This article explores the current capabilities of AI, surveys expert forecasts, and highlights reasons for caution to answer the questions: how far are we from AGI, is artificial general intelligence coming soon, and how close are we to developing general AI?

Progress and Limitations

Today’s AI tools achieve impressive results on specialised benchmarks. Microsoft’s GPT‑4 delivers near‑human performance on diverse tests in mathematics, coding, law, and medicine, while DeepMind’s Gemini system recently solved complex mathematics problems at an Olympiad level. According to the 2025 AI Index report, scores on demanding tasks like MMMU, GPQA, and SWE‑bench improved dramatically in a single year, and some AI agents now outperform humans in specific programming challenges. These gains reveal that advanced systems can learn to handle a wide variety of tasks under guided settings, suggesting a degree of emerging generality.

Despite rapid improvement, these tools remain narrow. They excel when given clear instructions and ample training data but falter without structured guidance or when tasks require physical interaction. Even the most capable models lack the manual dexterity and adaptability needed for jobs such as construction or food preparation. Some researchers argue that current approaches emphasising pattern recognition cannot capture the flexible, context‑rich memory and reasoning that define human cognition. Others warn that the hype surrounding AGI may divert resources away from more pressing technological and social challenges.

Predictions From Researchers and Forecasters

Industry leaders have shortened their timelines for AGI. Demis Hassabis of DeepMind recently suggested that we might see AGI within five to ten years. OpenAI’s Sam Altman has expressed excitement about major breakthroughs around 2025. Surveys of AI researchers reflect cautious optimism: a 2023 poll estimated a 25 % chance of high‑level machine intelligence in the early 2030s and a 50 % chance by the mid‑2040s. Metaculus forecasters consider a similar level of progress likely even sooner, with a 25 % chance of AGI by 2027 and an even shot by 2031. Tech entrepreneurs like Eric Schmidt, Elon Musk, Dario Amodei, and Jensen Huang have speculated that AGI could appear between 2026 and 2029.

Skeptical Voices

Other experts argue that AGI may be decades away or even unattainable. AI researcher François Chollet points out that many standard benchmarks can be solved through memorisation and that performance drops sharply when problems are rephrased or created after the model’s training data. A team led by cognitive scientist Iris van Rooij has argued that replicating human‑like cognition with current machine‑learning techniques is computationally implausible because human memory spans seconds to decades and relies on context and meaning in ways that machines struggle to emulate.

A Range of Forecasts

Estimates for when AGI might arrive vary widely. Past surveys place a 50 % chance of AGI anywhere from the late 2030s to the 2060s. Some experts believe it may never happen, while a minority think it could emerge as soon as this decade. Because definitions of AGI differ, and because predicting technological breakthroughs is inherently uncertain, these forecasts should be interpreted as informed speculation rather than firm timelines.

Here are several prominent outlooks:

• AI company leaders: 25 % chance of AGI by 2026; they have the closest view of cutting‑edge research but also strong incentives to generate excitement.

• Academic researchers (2023 survey): 25 % chance by around 2032, based on a definition that requires performing most tasks better than humans.

• Metaculus forecasters (early 2025): 25 % chance by 2027; this community emphasises quantitative forecasting but defines AGI broadly enough to include advanced robotics.

• Superforecasters (2022): 25 % chance by 2047; these predictions predate the ChatGPT boom and therefore may be conservative.

• Samotsvety (2023): 25 % chance by 2029; this group is known for strong forecasting records and an interest in AI progress.

  
  

How Close Are We?

AI systems exhibit remarkable capabilities, but an autonomous, self‑motivated intelligence that can perform any task as well as or better than humans does not yet exist. Current models lack curiosity, long‑term memory, and the ability to operate seamlessly in the physical world. Achieving AGI will likely require breakthroughs in reasoning, memory, agency, and alignment, not just scaling up existing neural networks. At the same time, leading organisations emphasise the need for responsible AI development to ensure safety, fairness, and human control.

Conclusion

The road to artificial general intelligence remains uncertain. While some experts see a significant likelihood that AGI could emerge within the next two decades, others caution that fundamental hurdles may delay or prevent its arrival. For individuals and organisations, the prudent approach is to embrace advanced AI tools as they evolve, invest in critical AI literacy, and prioritise ethical safeguards. Businesses focused on security and preparedness can gain an edge through strategic intelligence and security research, surveillance and field investigations, and digital cyber investigations. Whether AGI appears soon or remains a distant goal, thoughtful engagement and responsible development will shape how this transformative technology impacts society.

=======

About the Author:

Dr. Oludare Ogunlana is a Professor of Cybersecurity and an AI researcher. As an academic practitioner, he explores the intersection of emerging technologies, digital investigations, and national security through research, teaching, and strategic consulting.