California: At the recent Google I/O developer conference held in California, Demis Hassabis, Chief Executive Officer of Google DeepMind, indicated that Artificial General Intelligence (AGI) could become a reality as early as 2029. His statement has reignited global debate on the pace of artificial intelligence development and the preparedness of governments and societies for such a transformative shift.
Hassabis, a leading voice in AI research, remarked that humanity is currently in the “foothills of the singularity”, suggesting that exponential technological progress is approaching faster than many had anticipated. While he had earlier projected AGI to emerge around 2030, he now believes that advancements in AI systems, particularly intelligent agents, could accelerate this timeline by at least a year.
Growing urgency around AGI preparedness
Speaking after his keynote session, Hassabis emphasised the need for urgency in addressing the societal implications of AGI. He acknowledged that discussions about AI’s long-term impact are largely confined to technology circles, while its consequences could extend to nearly every sector, including healthcare, education, employment, and governance.
He noted that governments, economists, and policymakers must move faster in preparing frameworks that can manage both the opportunities and risks associated with AGI. According to him, the use of strong language was deliberate, aimed at pushing stakeholders to recognise the scale of change that could unfold within a few years.
The rapid development of AI-powered agents — systems capable of executing tasks independently — has further reinforced his outlook. Hassabis pointed out that current models are already demonstrating early signs of what he termed “soft self-improvement”, as they significantly enhance the productivity of human developers and engineers.
Self-improving AI remains a key milestone
One of the most critical benchmarks in the journey towards AGI is the concept of recursive self-improvement. This refers to AI systems that can enhance their own capabilities without human intervention, potentially leading to a cycle of accelerating intelligence growth.
Hassabis clarified that while existing AI systems are not yet capable of fully independent self-improvement, research laboratories across the world are closely monitoring this development. He stressed that the pace of innovation in this area is clearly accelerating, making it a focal point for leading AI organisations.
Experts believe that achieving this milestone would mark a significant leap towards AGI, as it would enable machines to evolve beyond their initial programming and continuously refine their performance.
Lack of consensus on AGI definition
Despite growing interest and investment in AGI, there remains no universally accepted definition of the term. Different leaders in the AI industry interpret AGI in varying ways, leading to ongoing debate.
OpenAI CEO Sam Altman has described AGI as a “weakly defined term”, though his organisation broadly considers it to be a system capable of solving complex problems at a human level across multiple domains. Meanwhile, Anthropic CEO Dario Amodei has expressed reservations about the term itself, preferring to use “powerful AI” to describe future systems.
This lack of consensus complicates efforts to measure progress and set regulatory standards. However, it also reflects the evolving nature of AI technology and the difficulty of defining intelligence in precise terms.
A scientific benchmark for true AGI
Hassabis proposed a unique benchmark to determine whether AGI has truly been achieved. He suggested that a genuine AGI system should be capable of independently discovering new scientific theories.
To illustrate this, he presented a hypothetical test: training an AI model only on knowledge available up to 1911 and then assessing whether it could develop the theory of general relativity, as Albert Einstein did in 1915. Such an achievement, he argued, would demonstrate true general intelligence rather than mere pattern recognition or data-driven prediction.
This benchmark highlights the distinction between current AI systems, which primarily rely on existing data, and future systems that could generate entirely new knowledge.
Industry trends point to accelerated timelines
Across the AI industry, there is a noticeable shift towards shorter timelines for AGI development. While earlier predictions often placed AGI decades away, recent statements from industry leaders suggest that it could arrive within the next decade.
In 2024, Dario Amodei had predicted that powerful AI systems might emerge by 2026, although this timeline has not yet materialised. Nevertheless, the broader trend indicates increasing confidence that AGI is approaching sooner rather than later.
The rapid evolution of large language models, improved computational power, and growing investment in AI research are all contributing factors to this shift.
Conclusion
Demis Hassabis’ revised timeline for AGI underscores the accelerating pace of artificial intelligence development and the urgent need for global preparedness. While uncertainty remains around the exact definition and arrival of AGI, the consensus among leading experts is that transformative changes are imminent.
As the world moves closer to this technological milestone, proactive planning, ethical considerations, and regulatory frameworks will play a crucial role in ensuring that the benefits of AGI are maximised while potential risks are effectively managed.