The swift progress in artificial intelligence (AI) has introduced profound ethical dilemmas within AI laboratories, prompting urgent discussions about safety protocols, bias mitigation, and accountability. A report commissioned by the U.S. State Department and authored by Gladstone AI employees highlights significant safety concerns among workers at leading AI labs such as OpenAI, Google DeepMind, Meta, and Anthropic. The report underscores fears that these labs, driven by the pursuit of powerful AI systems, may possess insufficient safety protocols to prevent artificial general intelligence (AGI) from escaping control. Employees also expressed concerns about cybersecurity vulnerabilities and the potential success of intellectual property exfiltration attempts. Some lab workers, wary of the risks associated with whistleblowing, worry that internal decision-making and management incentives may inadequately address catastrophic risks. The report suggests that governmental oversight and rigorous standards are necessary to mitigate these risks and criticizes the current reliance on easily manipulated AI evaluations. time.com
In addition to safety concerns, ethical lapses in AI labs have been linked to biases in AI systems. Rumman Chowdhury, an AI scholar and co-founder of Humane Intelligence, warns that advances in AI are at risk due to the lack of ethical considerations in its development. She points out that while much attention is given to potential long-term existential risks of AI, real and present harms are being overlooked. For instance, current AI systems have wrongly imprisoned individuals and denied them necessary organ transplants. Chowdhury also criticized the tech industry's persistent exclusion of diverse perspectives, which she links to a resurgence of "bro culture," reminiscent of the tech industry's climate in 2017. axios.com