Ethical Machines

Everyone knows biased or discriminatory AI bad and we need to get rid of it, right? Well, not so fast.

I’m bringing one of the best episodes from Season 1 back. I talk to David Danks, a professor of data science and philosophy at UCSD. He and his research team argue that we need to reconceive our approach to biased AI. In some cases, David thinks, it can be beneficial. Good policy - both corporate and regulatory - needs to take this into account.

Data about us is collected, aggregated, and shared in more ways than we can count. In some cases, this leads to great benefits. In others, a great deal of harm. But at the end of the day, the truth is that it’s all out of control. No individual, nor any private company, nor any government has a grip on what gets collected, what gets done with it, and what the societal impacts are. In this episode I talk to Aram Sinnreich and Jesse Gilbert about their new book, “The Secret Life of Data,” in which they explain the complexity and how we should begin to take back control.

With the ocean of social media content we need AI to identify and remove inappropriate material; humans just can’t keep up. But AI doesn’t assess content the same way we do. It’s not a deliberative body akin to the Supreme Court. But because we think of content moderation as a reflection of human evaluation, we then make unreasonable demands of social media companies and ask for regulations that won’t protect anyone. When we reframe what AI content moderation is and has to be, my guest argues, that leads us to make more reasonable and more effective demands of social media companies and government.

AI + nuclear capacities sounds like a recipe for disaster. Some people think it could cause mass extinction. While it’s easy to let our imaginations run wild, insight into how the military actually incorporates AI into its weapons and operations is a much better idea. Heather gives us precisely those insights and (thus) the opportunity to think clearly about the threat.

AI holds a lot of promise in making faster, more accurate diagnoses of our ailments. But if they are too influential, might they undermine our doctors’ ability to understand the rationale for the diagnose? And could it undermine the aspect of the doctor-patient relationship that is crucial for maintaining our patient autonomy?

Privacy is important. But I think we mostly misconceive the nature of privacy and data privacy. I argue we should rethink data privacy so that we can both focus better on how to protect people and so we can enable legitimately desirable innovations.

Technologist’s are racing to create AGI, artificial general intelligence. They also say we must align the AGI’s moral values with our own. But Professors Ariela Tubert and Justin Tiehen argue that’s impossible. Once you create an AGI, they say, you also give them the intellectual capacity needed for freedom, including the freedom to reject your given values.

Of course, decreasing racial disparities in healthcare is ethically imperative. But does it sometimes require too great a moral sacrifice? If it costs more lives than an non-equitable distribution of healthcare resources, should we really do it? Professors Guha Krishnamurthi and Eric Vogelstein argue that equity is not always a moral trump card.

#ai

Could online sexual assault be as morally bad as in-person sexual assault? Honestly, that initially struck me as a bit crazy. But Professor John Danaher makes some very compelling arguments.

Welcome to Ethical Machines—a weekly podcast from author and ethicist Reid Blackman. Episodes drop June 27th.