Confidence calibration in LLMs, i.e., aligning their self-assessed confidence with the actual accuracy of their responses, enabling them to self-evaluate the correctness of their outputs. However, current calibration methods for LLMs typically estimate two scalars to represent overall response confidence and correctness, which is inadequate for long-form generation where the response includes multiple atomic facts and may be partially confident and correct. These methods also overlook the relevance of each fact to the query. To address these challenges, we propose a Fact-Level Calibration framework that operates at a finer granularity, calibrating confidence to relevance-weighted correctness at the fact level. Furthermore, comprehensive analysis under the framework inspired the development of Confidence-Guided Fact-level Self-Correction ($\textbf{ConFix}$), which uses high-confidence facts within a response as additional knowledge to improve low-confidence ones. Extensive experiments across four datasets and six models demonstrate that ConFix effectively mitigates hallucinations without requiring external knowledge sources such as retrieval systems.

In the winter of 1958, a 30-year-old psychologist named Frank Rosenblatt was en route from Cornell University to the Office of Naval Research in Washington DC when he stopped for coffee with a journalist. Rosenblatt had unveiled a remarkable invention that, in the nascent days of computing, created quite a stir. It was, he declared, "the first machine which is capable of having an original idea". Rosenblatt's brainchild was the Perceptron, a program inspired by human neurons that ran on a state-of-the-art computer: a five-tonne IBM mainframe the size of a wall. Feed the Perceptron a pile of punch cards and it could learn to distinguish those marked on the left from those marked on the right.

In May of 1997, Garry Kasparov sat down at a chess board in a Manhattan skyscraper. Kasparov, considered the best chess player of all time, wasn't challenging another grandmaster. He was playing with an AI called Deep Blue. Deep Blue was one of the world's most powerful supercomputers, built by IBM with a specific goal in mind: to beat humanity at its own game. For IBM, billions of dollars worth of business clout was on the table, and to a certain extent, Kasparov was playing for the fate of chess itself. He had never lost a multi-game match in his entire career. Could a machine beat him? Newsweek ran a cover story with his picture alongside the words "The Brain's Last Stand." As Kasparov joked years later, "No pressure." Thanks to ChatGPT, once hypothetical questions about the future of work, art, and disinformation are now immediate concerns.

ECG's radio show and podcast, Healthcare Upside Down, offers unfiltered perspectives on what's working in US healthcare and what's not. Hosted by ECG principal Dr. Nick van Terheyden, each episode features guest panelists who explore the upsides and downsides of healthcare in the US--and how to make the system work for everyone. Early computing was based on programming languages that incorporated simple logic statements: if this happens, then do this; otherwise, do that. But the techniques and capabilities have moved far beyond that, and we now have high-level tools that can ingest large amounts of content and pull it together into some proxy of knowledge. Our guest on episode 68 of Healthcare Upside Down is John Lee, MD, emergency physician and digitician.

It's no secret that artificial intelligence now plays a big role in Chess. Chess engines, such as LeelaZero and AlphaZero took the world by storm and now pose danger to the world's best -- Stockfish. The Question is: When and how did the AI become so strong? In the Garry Kasparov era, humans were dominating the Chess world. Karpov, Kasparov, Ivanchuk, Gelfand…They were all at the top of the mountain.

AI has done remarkably well in performing specific functions with the level of accuracy it displays. But does that mean it has come close to human intelligence? I will begin by acknowledging Ragnar Fjelland, whose paper forms the primary inspiration for this article. The three milestones led to the impression that an AI closer to human intelligence is just "around the corner." First Milestone: The first milestone, according to Fjelland, in AI research is IBM's chess-playing computer Deep Blue.

By Aleksandras Šulženko Image credit: 123RF (with modifications) Imagine you are about to sit down to play a game with a friend. But this isn’t just any friend - it’s a computer program that doesn’t know the rules of the game. It does, however, understand that it has a goal, and that goal is to…

Growing up in north London, the child of a Greek Cypriot father and a Chinese Singaporean mother, Hassabis was a child prodigy in chess from the age of 4. He began writing his own computer games at 8, created one of the first video games to use AI at 17, and founded his own video game company not long after graduating from Cambridge University at 20. So perhaps it makes sense that Hassabis's AI startup DeepMind, founded in 2010 and sold to Google just four years later, would achieve its first major successes with AI models that used deep reinforcement learning to rapidly master video games like Space Invaders and Q*bert without any knowledge of the actual rules. That was followed with AlphaGo, which learned the ancient strategy board game of Go and would in 2017 defeat the world's number one human player -- an event that did perhaps more than anything else to awaken the world to the rapid progress of AI. New models could dominate a variety of games even faster, reducing the time and human intervention needed to acquire mastery.

Growing up in north London, the child of a Greek Cypriot father and a Chinese Singaporean mother, Hassabis was a child prodigy in chess from the age of 4. He began writing his own computer games at 8, created one of the first video games to use AI at 17, and founded his own video game company not long after graduating from Cambridge University at 20. So perhaps it makes sense that Hassabis's AI startup DeepMind, founded in 2010 and sold to Google just four years later, would achieve its first major successes with AI models that used deep reinforcement learning to rapidly master video games like Space Invaders and Q*bert without any knowledge of the actual rules. That was followed with AlphaGo, which learned the ancient strategy board game of Go and would in 2017 defeat the world's number one human player -- an event that did perhaps more than anything else to awaken the world to the rapid progress of AI. New models could dominate a variety of games even faster, reducing the time and human intervention needed to acquire mastery.

Either it is a master or a servant. And there are two kinds of relationships that humanity employs with technology. Either it fears, is destroyed, and is overwhelmed by technological progress, or it dictates this progress, drives it forward, and uses it. And then there are two major social constructs that keep the members of society in compliance with the prevailing ideology of life. These are religion and the state.