Shriyash Kaustubh Upadhyay

To boost both safety and innovation, regulators should mandate that large AI laboratories release small, openly accessible "analog models"—scaled-down versions trained similarly to and distilled from their largest proprietary models.