Just before 3 a.m. one night this month, Scott Slumpff was awakened by the ding of a text message. "An ALERTCalifornia anomaly has been confirmed in your area of interest," the message said. Slumpff, a battalion chief with the California Department of Forestry and Fire Protection, sprang into action. The message meant the agency's new artificial intelligence system had identified signs of a wildfire with a remote mountaintop camera in San Diego County. Within minutes, crews were dispatched to the burgeoning blaze on Mount Laguna -- squelching it before it grew any larger than a 10-foot-by-10-foot spot.

Smoke from the Airport Fire seen from the nearby town of Big Pine, California. The nation's most wildfire-prone state is teaming up with an artificial intelligence platform that "never sleeps" and can detect potential fires before they spiral into chaos. The California Department of Forestry and Fire Protection (Cal Fire) is partnering with University of California San Diego's ALERTCalifornia, a public safety program that studies natural disasters, to test a $24 million AI program. "We've got an automated system that never sleeps, never rests, watching the North Bay 24 hours, seven days a week," Cal Fire Napa-Lake-Sonoma Unit Chief Mike Marcucci told Fox 2. California is the state most threatened by wildfires in the nation, with 7,396 recorded wildfires in 2021 alone, and 2.5 million acres burned. The Golden State recorded another 7,447 wildfires last year, which burned a combined 331,360 acres.

Large language models (LLMs) transfer well to new tasks out-of-the-box simply given a natural language prompt that demonstrates how to perform the task and no additional training. Prompting is a brittle process wherein small modifications to the prompt can cause large variations in the model predictions, and therefore significant effort is dedicated towards designing a painstakingly "perfect prompt" for a task. To mitigate the high degree of effort involved in prompt-design, we instead ask whether producing multiple effective, yet imperfect, prompts and aggregating them can lead to a high quality prompting strategy. Our observations motivate our proposed prompting method, ASK ME ANYTHING (AMA). We first develop an understanding of the effective prompt formats, finding that question-answering (QA) prompts, which encourage open-ended generation ("Who went to the park?") tend to outperform those that restrict the model outputs ("John went to the park. Output True or False."). Our approach recursively uses the LLM itself to transform task inputs to the effective QA format. We apply the collected prompts to obtain several noisy votes for the input's true label. We find that the prompts can have very different accuracies and complex dependencies and thus propose to use weak supervision, a procedure for combining the noisy predictions, to produce the final predictions for the inputs. We evaluate AMA across open-source model families (e.g., EleutherAI, BLOOM, OPT, and T0) and model sizes (125M-175B parameters), demonstrating an average performance lift of 10.2% over the few-shot baseline. This simple strategy enables the open-source GPT-J-6B model to match and exceed the performance of few-shot GPT3-175B on 15 of 20 popular benchmarks. Averaged across these tasks, the GPT-J-6B model outperforms few-shot GPT3-175B. We release our code here: https://github.com/HazyResearch/ama_prompting

Britain recorded its hottest day on record last Tuesday, sparking a series of wildfires across the UK. Firefighters in London described the blazes tearing through homes and buildings as'absolute hell' after receiving 1,600 calls for assistance. They warned the public that wildfires are likely to break out every three years, and that the destruction of homes should be a'wake-up call' to the country. Global experts are urging countries to reach their net-zero targets and halt climate change. This is widely agreed to be the leading cause of the recent spate of wildfires, as rising temperatures evaporate more moisture from the ground, drying out the soil and making vegetation more flammable if sparked.