In a new review of artificial intelligence development, AI expert Dr Roman V. Yampolskiy raised significant concerns regarding the safety and control of AI systems.

According to him, there is no current evidence to suggest that AI can be controlled safely, posing existential risks to humanity if not addressed properly.

In his extensive review, “AI: Unexplainable, Unpredictable, Uncontrollable,” Dr Yampolskiy looks into the potential impact of AI on society, highlighting the need for stringent safety measures in its development.

He says that AI has the potential to bring about significant advancements, but its uncontrolled development could lead to catastrophic consequences. 

According to him, one of the fundamental challenges in controlling AI lies in its ability to learn, adapt, and act semi-autonomously in novel situations. As AI systems become more capable, the range of possible decisions and failures becomes infinite, making it virtually impossible to predict and mitigate all safety issues.

“There seems to be a lack of published evidence to conclude that a less intelligent agent can indefinitely maintain control over a more intelligent agent. As we develop intelligent systems that are less intelligent than we are, we can maintain control, but once such systems become more intelligent than we are, we lose that ability,” he said. 

Moreover, AI’s inability to explain its decisions poses a significant obstacle to ensuring transparency and accountability in its operation. As AI systems are increasingly deployed in critical domains such as healthcare, finance, and security, the need for explainable AI becomes paramount to prevent bias and ensure trustworthiness.

Recommended reading

• UK Gov Announces £100 Million for AI Regulation & Research
• Where Will Companies Look to For AI Regulation Guidance?
• Report: Business Leaders Want Sector-specific AI Regulation

He argues that humanity must strike a balance between AI capability and control to minimise the potential for catastrophic outcomes.House of Lords Warns UK Could Miss Out on AI Goldrush

To address the challenges posed by uncontrolled AI development, Dr Yampolskiy proposes several measures, including the categorisation of AI systems as controllable or uncontrollable. He advocates for increased funding and research efforts in AI safety and security to mitigate risks and ensure the responsible development of AI technology.

While achieving completely safe AI may be challenging, Dr Yampolskiy believes that concerted efforts can make AI safer in proportion to the level of investment and research. He urges policymakers and researchers to seize the opportunity to address AI safety concerns before they escalate into existential threats.

This is on the backdrop of the European Union’s 27 member states approving the guidelines for the world’s first AI guidelines rulebook and regulations. The move by the EU is just one step forward in not just regulating, but understanding how AI will function in our modern society. 

The UK has yet to introduce any such regulations yet. While the UK lead the AI safety summit last autumn, their non-statutory approach to AI legislation has been critiqued by the house of lords as well as AI developers.