The dawn of Artificial General Intelligence (machines that can match or exceed human intelligence across all areas of knowledge) is coming closer. Predictions vary, but we are nearing the tipping point.
When AGI appears, there will be significant change in society, medicine, technology, business and everyday life. Can we predict a date for its arrival?
Ray Kurzweil* predicts that AGI will be achieved by 2029 and the technological singularity will occur by 2045.
MIT Technology Review’s report The Road to Artificial General Intelligence, produced with Arm, reports that the timeline for AGI once thought to be decades away is now measured in years. Forecasts give at least a 50% chance of AGI-level systems by 2028.
Anthropic’s Dario Amodei foresees ‘powerful AI’ emerging as early as 2026, with Nobel-level expertise and autonomy to reason toward goals. Sam Altman of OpenAI believes AGI is already visible, comparing its social impact to electricity or the internet.
However, true general intelligence remains elusive. Today’s models can write code and discover drugs but still stumble over puzzles a child could solve. As François Chollet of Ndea explains, intelligence is not memorisation but the ability to recombine knowledge into new patterns. On this measure, even the largest language models score close to zero.
Compute, architecture and adaptability
Progress depends on three interlinked forces: algorithms, hardware and orchestration.
Before 2010, AI compute (computational resources and processing power) needs doubled roughly every 21 months. Now that we have LLMs (Large Language Models such as ChatGPT), that interval has dropped to just under six months. AGI may demand more than 10^16 teraflops, orders of magnitude beyond today’s capacity. This means that efficiency has become as important as power.
Ian Bratt of Arm sees a future where intelligence runs at every layer of the digital stack, from hand-held devices to the cloud. Heterogeneous computing using CPUs, GPUs, neural and tensor processors together, offers a route forward.
Equally, AGI’s advance relies on improved software orchestration. Developers must be able to distribute workloads across these mixed environments without rewriting code and new standards are emerging to manage such dynamic systems. The compute stack must therefore evolve holistically, combining hardware innovation with smarter algorithms and energy-aware infrastructure.
Beyond human benchmarks
One of the most interesting insights from MIT is that measuring machine intelligence against human cognitive capability may be not be helpful. Rumman Chowdhury of Humane Intelligence argues that we should look for adaptability, social understanding and environmental awareness, forms of intelligence found across biological systems. This shift reframes AGI not as a replication of human thought, but as the emergence of diverse, complementary intelligences.
The race toward the AI singularity
AI Multiple’s review of AGI forecasts suggests that researchers now see a realistic chance of AGI within five years. A Metaculus consensus model puts the 50% probability around 2030. The driving factors include cheaper compute, scaling laws that continue to hold and accelerating hardware efficiency.
However, as Popular Mechanics notes, the so-called singularity, the point at which machines improve themselves beyond human control, could arrive suddenly, perhaps within a single research cycle. Some experts warn that this could happen within months once a system gains the capacity for recursive self-improvement. Others counter that alignment, safety frameworks and regulatory oversight will act as natural brakes.
An inflection point
The trajectory toward AGI mirrors earlier technological revolutions: unpredictable, uneven, yet unstoppable once critical thresholds are reached. What distinguishes this era is the convergence of massive data, near-limitless compute and a global race to build adaptable, multimodal systems. Whether AGI arrives in three years or thirty, the preparation must start now with attention not only on power and capability but also to ethics, safety, governance and human benefit.
In the words of MIT’s report, intelligence may ultimately prove to be efficiency: the ability to solve new problems with limited resources. By that definition, the journey to AGI is as much about learning to think differently as it is about teaching machines to think like us.
Sources:
> MIT Technology Review / Arm report
> When will AGI/Singularity happen?
> Humanity may achieve the singularity within the next three months.
* Ray Kurzweil, author of The The Singularity is Near (2005) and The Singularity is Nearer (2023).
The Singularity is a point in time when technological growth, driven by AI, accelerates so fast that it transforms civilization.
If you are interested in advertising within this website, visit the Contact page.
