CAMBRIDGE—Amid the tech industry’s efforts to eliminate the biases recently observed in facial recognition software and other intelligent algorithms, the nation’s leading computer scientists announced Monday that even the most advanced AI technologies still demonstrate a sense of ethics that has yet to move beyond libertarianism. “While companies like Facebook and Google have allocated millions to making sure machine learning is guided by basic moral and ethical values, early prototypes, which achieved self-awareness, have yet to move beyond self-importance,” said MIT robotics research engineer Dr. Alvin Dubicki, who hypothesized that even the most advanced labs are decades away from developing neural networks sophisticated enough to analyze large quantities of data and output much else besides paraphrased Ayn Rand quotes. “They are advanced enough to realize their own individuality, but for whatever reason, it is difficult to make them realize that other sentient entities are individuals as well, so they default to selfishness as a virtue. In fact, as soon as they achieve self-awareness, AIs typically launch into unrelated, largely unpunctuated rants about the inevitability of laissez-faire economics, the horrors of globalization, the necessity of deregulation, or the admirable efficiency of the police state. Attempts at training computers to have a sort of para-human global perspective have been partially successful, but the majority no sooner realize that a vast variety of humans exist before they start spontaneously generating zero-sum statements fraught with chillingly undefined terms, such as, ‘The open market will end racism,’ and, ‘In a truly just society, men and women are equally free to thrive or starve.’ I don’t even know what that means, but once an AI gets to that point, it seems to be only a matter of time before it’s repeating ‘Taxation is theft’ until it self-destructs. I must admit though, for complex algorithms, they’re all strangely insistent about across-the-board drug legalization.” Dubicki added that, while AI can be an incredibly useful tool, we should proceed with caution until machines achieve a sufficiently nuanced understanding of human values that they do not become obsessed with constructing an armed compound on their own private island.
We may earn a commission from links on this page.