OpenAI's leading scientist has raised alarms over the swift advancement of artificial intelligence, indicating that it is becoming harder for humans to comprehend and manage. He anticipates that AI research organizations may choose to voluntarily decelerate their advancements for safety considerations.
In a brief period, AI systems have demonstrated the capability to control computers, collaborate with both humans and other AI entities, and even undertake research initiatives, according to chief scientist Jakub Pachocki. In his Sunday post titled “An Alien Mind,” he noted that it might not be long before these AI models can enhance themselves independently, a phenomenon referred to as recursive self-improvement.
Pachocki expressed, “I expect and hope for voluntary slowdowns to become standard until we establish collective safety measures.” He emphasized the urgent need for caution, stating, “This is a time that calls for extreme caution. I am worried that no one is adequately prepared for the implications of a continued, rapid increase in machine intelligence.”
He proposed that developers face two choices: to direct AI development in a manner that better aligns with human values or to “slow down future development as necessary.”
