- Report warns superintelligent AI could exceed human abilities within five years
- Loss of meaningful human control may occur if AI strategies outpace institutions
- AI poses risks including chemical, biological threats, and democratic integrity harm
A new report from the UK-based Alan Turing Institute has warned that there was a 'realistic possibility' that humans might lose control of 'superintelligent' artificial intelligence (AI) machines within five years. Titled "Frontier AI poses credible near-term risks -- demanding practical research, robust regulation, whole-system verification and international cooperation," the report published on Wednesday (Sep 23) talks about AI exceeding human capabilities, leading to catastrophic risks.
The institute set up to manage AI developments in Britain warned that there was a risk that humans will increasingly hand control of critical tasks to the technology as it improves itself before pushing humans out of the loop.
“There is a realistic possibility that ‘superintelligent' AI systems will emerge in the next five years, which exceed human abilities across almost all important cognitive tasks,” the report highlighted.
The report warned that if AI systems can devise strategies that institutions cannot adequately evaluate, then substantive human control could disappear.
"A critical threshold is reached when preserving meaningful human authority requires accepting slower decisions or lower measured task performance. Human control is meaningful only when decision-makers have a reliable basis for assessing the evidence, challenging the options and enforcing an alternative," it said.
"If AI systems devise strategies that institutions cannot adequately evaluate and act faster than organisations can respond, substantive human control could disappear."
The report also talked about AI posing chemical and biological threats, in addition to undermining the integrity of both our information environment and wider democratic system.
Also Read | 'Like An Idiot, I Believed Them': Woman Fired On First Day Of New Job For Being 'Honest'
Pause Frontier AI Development
The report comes in the backdrop of a recent string of doomsday warnings from top AI researchers about pausing frontier AI development before it slips beyond human control. Anthropic CEO Dario Amodei proposed not only hitting the pause button but also embedding third-party evaluators, common safety benchmarks and international agreements with China to keep the technology in check.
Former Anthropic researcher Jacob Coxon warned AI could kill us all by the end of the decade. Evan Hubinger, Anthropic's alignment science lead, echoed Coxon's warning, saying there was more than 10 per cent chance of such an event within the next decade.
Josh Engels, who worked on Google DeepMind's AGI safety team, has left the company and joined AI evaluation organisation METR. He said he made the move because he believes the stakes surrounding advanced AI have become too high.
The remarks show how quickly AI has advanced, from the hallucination-prone ChatGPT of 2022 towards what many see as a critical milestone: recursive self-improvement.
