OpenAI CEO warns of “superhuman persuasion” capabilities of AI systems
The CEO of OpenAI, Sam Altman, has issued a warning about the potential for AI systems to possess “superhuman persuasion” capabilities. Altman, whose company is behind the popular ChatGPT platform, made the statement on social media, stating that AI could excel in persuasion well before achieving general intelligence. He also warned that such capabilities could result in “very strange outcomes.”
These comments come as concerns about the potential capabilities of rapidly developing AI technology continue to grow, with some speculating that AI could surpass human cognitive functions. However, experts remain divided on the legitimacy of these fears.
While Altman did not provide specifics on what the “strange outcomes” might look like, some experts believe that fears of AI turning people into “mindless zombies” are unwarranted. Christopher Alexander, chief analytics officer of Pioneer Development Group, stated that persuasive AI will not manipulate people through subliminal messages. Instead, he argued that AI’s strength lies in its ability to identify persuasive content, based on machine learning and pattern recognition.
Alexander compared the potential effects of persuasive AI to those of digital advertising, where AI algorithms can identify what works best and at what frequency. He also noted that AI would not be able to outperform the mind-altering effects of social media, as they are already prevalent in society.
Aiden Buzzetti, president of the Bull Moose Project, cast doubt on the nearness of AI to achieving “superhuman persuasion” abilities. He pointed out that current platforms like ChatGPT still struggle with providing accurate information and often generate incorrect or imaginative responses. Buzzetti emphasized that AI has not yet reached the level of human intelligence and that fears of AI surpassing human cognitive abilities are misplaced.
Phil Siegel, founder of the Center for Advanced Preparedness and Threat Response Simulation (CAPTRS), argued that “we are already at that point” of superhuman persuasion with certain AI technologies. He warned that if a bad actor were to misuse data or draw incorrect conclusions, AI could convincingly persuade others of its accuracy. However, Siegel emphasized that the solution lies in questioning and not blindly accepting the expertise of either human or machine sources.
Jon Schweppe, policy director of American Principles Project, shared concerns about AI’s potential to deceive susceptible individuals and perpetrate fraud. In a lighthearted remark, Schweppe even joked that AI androids might one day run for Congress, fitting perfectly into the political landscape of Washington.
As the capabilities of AI continue to advance, ongoing discussions and debates around its potential benefits and risks are essential to managing its impact on society.