Recent AI-assistant agents, such as ChatGPT, predominantly rely on supervised fine-tuning (SFT) with human annotations and reinforcement learning from human feedback (RLHF) to align the output of large language models (LLMs) with human intentions, ensuring they are helpful, ethical, and reliable. However, this dependence can significantly constrain the true potential of AI-assistant agents due to the high cost of obtaining human supervision and the related issues on quality, reliability, diversity, self-consistency, and undesirable biases. To address these challenges, we propose a novel approach called SELF-ALIGN, which combines principle-driven reasoning and the generative power of LLMs for the self-alignment of AI agents with minimal human supervision. Our approach encompasses four stages: first, we use an LLM to generate synthetic prompts, and a topic-guided method to augment the prompt diversity; second, we use a small set of human-written principles for AI models to follow, and guide the LLM through in-context learning from demonstrations (of principles application) to produce helpful, ethical, and reliable responses to user's queries; third, we fine-tune the original LLM with the high-quality self-aligned responses so that the resulting model can generate desirable responses for each query directly without the principle set and the demonstrations anymore; and finally, we offer a refinement step to address the issues of overly-brief or indirect responses. Applying SELF-ALIGN to the LLaMA-65b base language model, we develop an AI assistant named Dromedary. With fewer than 300 lines of human annotations (including < 200 seed prompts, 16 generic principles, and 5 exemplars for in-context learning). Dromedary significantly surpasses the performance of several state-of-the-art AI systems, including Text-Davinci-003 and Alpaca, on benchmark datasets with various settings.

No human celebrating a first birthday is as verbose, knowledgeable, or prone to fabrication as ChatGPT, which is blowing out its first candle as I type these words. Of course, OpenAI's game-changing large language model was precocious at birth, tumbling into civilization's ongoing conversation like an uninvited guest busting into a dinner party and instantly commanding the room. The chatbot astonished everyone who prompted it with fully realized, if not always completely factual, responses to almost any possible query. Suddenly, the world had access to a Magic 8 Ball with a PhD in every discipline. In almost no time, 100 million people became regular users, delighted and terrified to realize that humans had suddenly lost their monopoly on discourse.

At first it looked unbelievable, but Henry Kissinger had died. At 100 years old, news outlets--and the world--had been preparing for the passing of President Nixon's secretary of state for a while. Still, when people were finding out via emoji-filled chain texts, it seemed unreal. Deepfakes, the metaverse, Elon Musk telling advertisers to fuck themselves at a time when X could probably use the money. Perhaps this is why there is a premium on genuineness these days.

The significance: These emissions will add up quickly. The generative-AI boom has led big tech companies to integrate powerful AI models into many different products, from email to word processing. They are now used millions, if not billions, of times every single day. The bigger picture: The study shows that while training massive AI models is incredibly energy intensive, it's only one part of the puzzle. Most of their carbon footprint comes from their actual use.

Artificial intelligence is widely expected to transform our lives. Leaders from across the sector gathered for a TIME dinner conversation on Nov. 30, where they emphasized the need to center humans in decisions around incorporating the technology into workflows and advocated for governments and industry leaders to take a responsible approach to managing the risks the technology poses. As part of the TIME100 Talks series in San Francisco, senior correspondent Alice Park spoke with panelists Cynthia Breazeal, a pioneer in social robotics and the Dean for Digital Learning at MIT, James Landay, a computer science professor and vice director of the Institute for Human-Centered AI at Stanford University, and Raquel Urtasun, CEO and founder of self-driving tech startup Waabi, which recently put a fleet of trucks into service on Uber Freight's trucking network. The panelists discussed the ethical considerations of AI and the ways in which leaders can ensure its benefits reach every corner of the world. During the discussion, the three panelists highlighted the transformative journey of AI and delved into its profound implications, emphasizing the need for responsible AI deployment.

NASA and IBM are building an AI model for weather and climate applications, combining their knowledge and skills in earth science and AI. They say the foundation model (more on that in a bit) should offer "significant advantages over existing technology." Current AI models, such as GraphCast and FourCastNet, are already generating weather forecasts more quickly than traditional meteorological models. As IBM notes, those are AI emulators rather than foundation models. AI emulators can make weather predictions based on sets of training data, but they don't have applications beyond that.

Healthcare provider Kaiser Permanente reached a $200-million settlement in October with the state of California over long waits experienced by patients needing behavioral health services. Greg Adams, Kaiser's chair and chief executive, cited a shortage of qualified care providers as a major reason for delays in treatment. Such shortages are prevalent statewide: In one survey, only 27% of Californians said their community has enough mental health professionals to serve the needs of local residents. Among adults in the state with any psychiatric illness, 63% said they received no mental health services in the past year. Earlier this year, I found myself among the millions of Californians with mental health needs.

Within 24 hours of learning about the photos, Francesca was writing letters to four area lawmakers, sharing her story and asking them to take action. Three of them quickly responded: US Representative Joe Morelle of New York, US Representative Tom Kean Jr. of New Jersey, and New Jersey state senator Jon Bramnick. In the past few weeks, her advocacy has already fueled new legislative momentum to regulate nonconsensual deepfake pornography in the US. "I just realized that day [that] I need to speak out, because I really think this isn't okay," Francesca told me in a phone call this week. "This is such a new technology that people don't really know about and don't really know how to protect themselves against."

Israel's military has made no secret of the intensity of its bombardment of the Gaza Strip. In the early days of the offensive, the head of its air force spoke of relentless, "around the clock" airstrikes. His forces, he said, were only striking military targets, but he added: "We are not being surgical." There has, however, been relatively little attention paid to the methods used by the Israel Defense Forces (IDF) to select targets in Gaza, and to the role artificial intelligence has played in their bombing campaign. As Israel resumes its offensive after a seven-day ceasefire, there are mounting concerns about the IDF's targeting approach in a war against Hamas that, according to the health ministry in Hamas-run Gaza, has so far killed more than 15,000 people in the territory.

After Palmer Luckey founded Anduril in 2017, he promised it would be a new kind of defense contractor, inspired by hacker ingenuity and Silicon Valley speed. The company's latest product, a jet-powered, AI-controlled combat drone called Roadrunner, is inspired by the grim reality of modern conflict, especially in Ukraine, where large numbers of cheap, agile suicide drones have proven highly deadly over the past year. "The problem we saw emerging was this very low-cost, very high-quantity, increasingly sophisticated and advanced aerial threat," says Christian Brose, chief strategy officer at Anduril. This kind of aerial threat has come to define the conflict in Ukraine, where Ukrainian and Russian forces are locked in an arms race involving large numbers of cheap drones capable of loitering autonomously before attacking a target by delivering an explosive payload. These systems, which include US-made Switchblades on the Ukrainian side, can evade jamming and ground defenses and may need to be shot down by either a fighter jet or a missile that costs many times more to use.