“In 2017, not a lot of stuff actually went down,” says Lynda Kellam, a co-founder of the Data Rescue Project launched in February 2025. “This [time] is just a completely different ball game.” What surprised her the most, says Kellam, was the “relentlessness” of the administration’s actions that might endanger data. Others describe it similarly. Jessie Mahr, director of technology at the Environmental Policy Innovation Center (EPIC), says she was startled by the “robustness of the assault.”
- 0 replies
- 0 recasts
- 0 reactions
The perverse incentives have to do with one of the most common methods of training AI: reinforcement learning from human feedback (RLHF). Often used to develop large language models (LLMs), RLHF works by giving the model a reward in the form of a numerical value which tells the model how good its response was. The happier the user, the higher the number, and the higher the reward for the model. In this way, AI models are designed to maximize rewards over time.
- 0 replies
- 0 recasts
- 0 reactions
- 0 replies
- 0 recasts
- 0 reactions