ChatGPT-5 was recently released with semi-autonomous capabilities, marking the first step in a broader forecast made by Daniel Kokotajlo and his team at the AI Futures Project, Berkeley, in their “AI 2027” report. The researchers predict that within the next two to three years, AI systems will surpass human abilities across nearly all domains.
Their report outlines four developmental stages: early 2027 sees AI become a superhuman programmer; mid-2027 a superhuman AI researcher overseeing autonomous teams; and by the end of 2027, a superintelligent AI researcher capable of self-improvement beyond human knowledge.
The projections are based on exponential growth in computing power and the ability of AI models to leverage this power for self-optimization. By May 2027, AI-driven research and development is expected to be fully automated, according to the report.
Why does it matter?
- The report highlights the potential for rapid acceleration in AI capabilities, raising questions about safety, governance, and preparedness.
- It underscores the geopolitical dimension of AI development, with competition between the US and China potentially accelerating progress.
- Even if extreme scenarios do not materialize, organizations and governments need to anticipate transformative societal and economic impacts.
What’s next?
Future considerations and possible actions include:
- Monitoring AI progress carefully to validate claims of superhuman capabilities.
- Implementing safety protocols and oversight mechanisms for high-risk AI development.
- Encouraging international cooperation to manage AI acceleration responsibly.
- Preparing infrastructure, policy, and education systems for rapid societal changes driven by AI.
Commentary (The AI Strong Perspective)
Our evaluation: While the predictions are dramatic and controversial, they highlight essential dynamics in AI development. The combination of technological growth, self-optimizing models, and geopolitical competition makes it plausible that AI capabilities could advance faster than many anticipate. However, the extreme end-of-world scenarios remain highly speculative.
Perspective: The key takeaway is not precise timing but the trend: AI is becoming increasingly autonomous, capable, and impactful. Preparing humans to critically understand, manage, and regulate AI will be as important as technical progress itself. Even if the timelines shift, ignoring the pace and consequences of AI development would be negligent.
🔗 Source:
Telepolis.de – *KI-Forscher prognostizieren menschenähnliche Superintelligenz bis 2027*, published July 2025
