AI Safety Issues Highlight Need for More Care in Development
-
LAION, a popular AI training data set, found to contain thousands of images of child sexual abuse. The data has been taken down and will be scrubbed before being republished.
-
The competitive pressures in AI are leading companies to rush products to market without enough thought given to ethics and potential harms. Recent examples include issues with Bing Chat, ChatGPT medical advice, and bias in DALL-E images.
-
EU AI regulations with threat of fines may help encourage more ethical AI development, but progress will be slow.
-
Other AI updates Microsoft Copilot can now compose songs; Rite Aid banned from using facial recognition tech; Google brings Gemini models to more products.
-
Study shows AI models can propagate outdated and racist medical advice. Need to be vigilant about bias and accuracy in health AI.