We present Second Thoughts, a new learning paradigm that enables language models (LMs) to re-align with human values. By modeling the chain-of-edits between value-unaligned and value-aligned text, with LM fine-tuning and additional refinement through reinforcement learning, Second Thoughts not only achieves superior performance in three value alignment benchmark datasets but also shows strong human-value transfer learning ability in few-shot scenarios. The generated editing steps also offer better interpretability and ease for interactive error correction. Extensive human evaluations further confirm its effectiveness.

We present Second Thoughts, a new learning paradigm that enables language models (LMs) to re-align with human values. By modeling the chain-of-edits between value-unaligned and value-aligned text, with LM fine-tuning and additional refinement through reinforcement learning, Second Thoughts not only achieves superior performance in three value alignment benchmark datasets but also shows strong human-value transfer learning ability in few-shot scenarios. The generated editing steps also offer better interpretability and ease for interactive error correction. Extensive human evaluations further confirm its effectiveness.

We present Second Thought, a new learning paradigm that enables language models (LMs) to re-align with human values. By modeling the chain-of-edits between value-unaligned and value-aligned text, with LM fine-tuning and additional refinement through reinforcement learning, Second Thought not only achieves superior performance in three value alignment benchmark datasets but also shows strong human-value transfer learning ability in few-shot scenarios. The generated editing steps also offer better interpretability and ease for interactive error correction. Extensive human evaluations further confirm its effectiveness.

If I ask you to look at the picture below and list some of the objects you see in the frame, what will they be? I am sure you would have probably come up with a long-drawn list of the objects you saw without giving a second thought. It may include cars, different colorful billboards, vibrant stores or the bollards on the road or the tall buildings or the potted plants alongside the road, and many other things. If I ask you to describe the picture you saw in just a sentence, you will probably say, "It's the New York Times Square!" again without giving a second thought. These tasks were extremely easy for you as even a person with below-average intelligence and understanding or even a six-seven-year-old kid can accomplish the same.

Whether you're searching something on Google, assessing a mortgage rate, or applying for a job, much of our lives today is informed by artificial intelligence. While AI helps systems operate quickly, it's not perfect. Like humans, these technologies are only as good as the information they get. "On Second Thought" host Virginia Prescott speaks with Dr. Ayanna Howard. Dr. Ayanna Howard is chair of the School of Interactive Computing at Georgia Tech.

Next Insurance has begun formally marketing a portal designed to help independent agents quickly quote and sell its policies. Next said the system, dubbed "Next for Agents," had been in beta testing since April 2019, though more than 3,000 agents are already using the platform, including some from BTIS, London Underwriters and First Connect Insurance. The insurtech startup focuses on commercial insurance for smaller businessowners, and its policies zero in on areas including general liability, professional liability and commercial auto insurance. As envisioned, independent agents can quickly quote and sell those policies now either through Next for Agents or through other independent agent platforms. Next CEO Guy Goldstein has previously argued that agents were no longer a necessary part of the equation if the buying process became automated, particularly for commercial lines. "If computers can give you exactly what you need, why use an agent?,"

When we look at a selection of faces, our brains can single out the familiar ones with no effort at all. This smooth process comes so naturally that most people never give it a second thought. But someone who does give this phenomenon a second thought is Doris Tsao, a professor of biology and biological engineering at the California Institute of Technology in Pasadena. Over recent years, Prof. Tsao has conducted a range of experiments that have attempted to get to the bottom of facial perception. In earlier studies, Prof. Tsao and her colleagues used functional MRI scans to search for relevant brain areas in humans and other primates.