Many artificial intelligences are inherently opaque, which means that part or all of the underlying structure is concealed, either deliberately to protect private information, due to the complexity of the AI model, or both. This can be a problem in cases where individuals are harmed by the decisions made by AI but find no way to challenge them.
Explainable AI: Solutions and Challenges
Many researchers are seeking solutions and have unified around a concept called “explainable AI,” but this concept also has its challenges. Notably, there is no real agreement on what explainable AI is and how it can be achieved. So how can we deal with these black boxes? In this podcast, we attempt to find out.
Subscribe to Nature Briefing
Subscribe to Nature Briefing, a daily summary of science news, opinions, and analyses delivered to your email every business day.
Don’t Miss an Episode of the Podcast
Don’t miss an episode of the podcast. Subscribe to the Nature Podcast on Apple Podcasts, Google Podcasts, Spotify, or your favorite podcast app. There is also an RSS feed available for the Nature Podcast.
Related References
– ChatGPT is a black box: How can AI research break it?
– Accessing the black box of artificial intelligence
– The World Week in AI Safety: Strong efforts to enhance research
– If AI becomes conscious: Here’s how researchers will know
Job Openings
– W3 Professorship in Clinical-Molecular Cardiology at the University of Bonn
– 22 PhD Student/Academic Staff (75%) positions at the University of Potsdam
– Multiple positions for appointment at Heidelberg University in Germany
– 30 PhD positions and 2 Postdoctoral positions at the University of Potsdam
– Open position for Principal Investigator at the Guangzhou Institute of Biomedicine and Health in China
You have full access to this article through your institution. Download the PDF file.
Subscribe to Nature Briefing
A daily summary of science news, opinions, and essential analyses delivered to your email every business day.
Email Address
Yes! Sign me up to receive the daily email from Nature Briefing. I agree to have my information processed in accordance with Nature’s Privacy Policy and that of Springer Nature Limited.
Subscribe Now
Leave a Reply