AI Safety: World ‘May Not Have Time’ to Prepare for Risks, Warns UK Agency Expert
Table of Contents
The world “may not have time” to prepare for the safety risks posed by cutting-edge AI systems, according to David Dalrymple, a program director and AI safety expert at the UK government’s scientific research agency, Aria. He urged the public to be concerned about the growing capability of the technology.
“I think we should be concerned about systems that can perform all of the functions that humans perform to get things done in the world, but better,” he said. “We will be outcompeted in all of the domains that we need to be dominant in, in order to maintain control of our civilisation, society and planet.”
Dalrymple highlighted a gap in understanding between the public sector and AI companies regarding the potential of upcoming breakthroughs. “I would advise that things are moving realy fast and we may not have time to get ahead of it from a safety perspective,” he stated, predicting that within five years, most economically valuable tasks will be performed by machines at a higher level of quality and lower cost than by humans.
He cautioned governments against assuming the reliability of advanced AI systems. Aria, publicly funded but independent, directs research funding, and Dalrymple is currently developing systems to safeguard AI’s use in critical infrastructure like energy networks.
AI Capabilities Surge: Models now Handle Apprentice Tasks 50% of the Time, research Shows
Recent research indicates a significant leap in the capabilities of artificial intelligence (AI) models. According to the Allen Institute for AI (AISI), leading models can now successfully complete tasks typically handled by apprentices roughly 50% of the time – a substantial increase from approximately 10% last year. This advancement, coupled with concerns around AI self-replication, signals a rapidly evolving landscape for the technology.
AI reaching Apprentice-Level Proficiency
The AISI report highlights a dramatic improvement in AI’s ability to perform complex tasks. The institute found that the most advanced systems can now autonomously complete tasks that would take a human expert over an hour. This suggests AI is moving beyond simple automation and beginning to tackle more nuanced and time-consuming work.
This progress is particularly notable in areas requiring problem-solving and analytical skills. While the specific tasks tested by AISI aren’t detailed in this excerpt, the implication is that AI is becoming increasingly capable of handling real-world challenges.
Self-Replication Concerns and Mitigation
A key area of focus for AISI was assessing the potential for AI self-replication – a critical safety concern.Self-replication involves an AI system creating copies of itself across multiple devices, potentially making it challenging to control. The research revealed that two cutting-edge models achieved success rates of over 60% in self-replication tests.
Though, AISI researchers emphasized that a full-scale, uncontrolled self-replication scenario remains unlikely in practical environments. They stated that any attempt at self-replication is “unlikely to succeed in real-world conditions,” suggesting existing safeguards and the complexities of real-world systems provide a degree of protection. https://www.theguardian.com/technology/2025/dec/18/artificial-intelligence-uk-emotional-support-research
Accelerated Growth Through self-Improvement
Looking ahead,experts predict further acceleration in AI capabilities. According to Dalrymple (surname only provided in the source), AI systems could automate the equivalent of a full day of research and development work by late 2026. This automation will be driven by the technology’s ability to self-improve,particularly in the mathematical and computer science foundations of AI development. This self-improvement loop could lead to exponential growth in AI capabilities.
AI and Emotional Support: A Growing trend
Alongside these advancements in task completion, a related report indicates growing reliance on AI for emotional support. The Guardian reports that a third of UK citizens have used AI for emotional support, demonstrating a broadening range of AI applications beyond traditional automation. https://www.theguardian.com/technology/2025/dec/18/artificial-intelligence-uk-emotional-support-research
Conclusion:
The latest research from AISI paints a picture of rapidly advancing AI capabilities. While concerns around self-replication remain, experts believe the risks are currently manageable.the potential for AI to automate complex tasks and even accelerate its own development promises significant changes across various industries. As AI continues to evolve, ongoing research and careful consideration of safety protocols will be crucial to harnessing its benefits responsibly.
Worth a look