Former OpenAI Researcher's 2024 AGI Prediction Now Seems Realistic as 2027 Approaches

Artificial intelligence (AI) development in mid-2026 is advancing at a terrifying and almost incomprehensible pace. Two years ago, as the public was still awestruck by the initial capabilities of large language models (LLMs), a stark warning emerged from the heart of the technology industry itself. In mid-2024, a former OpenAI researcher released a manifesto exposing the realities behind leading AI laboratories. He predicted that AI systems would reach Artificial General Intelligence (AGI) by no later than 2027. The 165-page manifesto titled ‘Situational Awareness: The Decade Ahead’ was penned by Leopold Aschenbrenner. He is no mere observer. Aschenbrenner was a former member of OpenAI’s Superalignment team, a specialised division established by co-founder Ilya Sutskever. The division’s mission was to ensure future AI would not defect, go rogue, or threaten human existence. Now in 2026, the document once dismissed as ‘dystopian prophecy’ has become essential reading for tech executives, investors, and policymakers in Washington. As 2027 draws nearer, here is an in-depth analysis of the ‘Situational Awareness’ report and how its warnings are unfolding before our eyes. The first part of the essay focuses on Aschenbrenner’s approach termed ‘Counting the OOMs’ (Orders of Magnitude). He dissects the fact that the AI industry is not moving by chance but along a precise, mathematical, and highly predictable exponential growth trajectory. Many laypeople view the leap in intelligence from GPT-2 in 2019 to GPT-3 and then GPT-4 in 2023 as pure technological magic. Aschenbrenner debunks this notion. This dramatic improvement stems from a simple mathematical equation: massive increases in computational power combined with continuously refined algorithmic efficiency. He calculates that AI models become significantly smarter each year due to exponential growth in training data and GPU processing power, measured in orders of magnitude (tenfold increases). Projecting the last five years’ trend forward, he confidently asserts that the threshold of human-equivalent intelligence (AGI) will be definitively crossed by 2027. With OOM leaps accelerating under tech companies’ push, AGI by 2027 is no longer theoretical but a matter of months away. A widespread misconception is that once AGI is achieved, technological innovation will hit a ‘brick wall’, plateau, and slow down.