Whether it is distilling statistics or determining protein structures, the reach of artificial intelligence (AI) is rapidly extending to encompass all aspects of scientific research. Starting this week, Nature takes a deep dive into how AI is helping to reshape the scientific enterprise. In this week’s issue, we look at why researchers are so excited about the burgeoning technology — and we also probe the risks posed by AI-generated disinformation. Over the coming weeks, we will explore other aspects of how AI could transform science, and will bring all of our content together in an online resource.

The increasing frequency and severity of wildfires comes with an immediate and obvious cost in terms of devastation and threat to life. But the air pollution generated by these conflagrations also has an important effect. In this week’s issue, Rongbin Xu and his colleagues present an estimate of global human exposure to air pollution from landscape fires (dominated by wildfires, but also including planned or controlled open land fires) between 2000 and 2019. They estimate that landscape fires contributed to 3.6% of total annual exposure to ozone and 6.1% of total exposure to particulate matter less than 2.5 micrometres in diameter (PM2.5) for the period 2010–19, with each person on average experiencing 9.9 days of exposure to substantial fire-sourced pollution per year. The team concludes that the global population is increasingly being exposed to pollution from landscape fires, but that there is significant disparity in exposure levels, with people in low-income countries experiencing four times higher levels of fire-sourced PM2.5 and ozone than those in high-income nations.

In September 2015, 193 countries agreed to work towards 17 goals aimed at improving the lives of people around the world. From eliminating poverty and reducing hunger to tackling global warming and taking care of biodiversity, the Sustainable Development Goals have since taken their place in corporate plans and government policy. But the world is now halfway towards the ambitious deadline of 2030 given to meet the goals, and there is still a lot of work to do. In this week’s special issue, Nature takes stock of how far we have come, and examines what still needs to be done to make meeting these global targets a reality.

The cover shows a thermal image of leaf temperature in an Australian tropical forest. In this week’s issue, Christopher Doughty and his colleagues reveal that a small percentage of leaves in tropical forest tree canopies might be approaching a critical temperature of 46.7 °C, above which photosynthesis begins to fail. The researchers estimated peak temperatures for tree canopies using high-resolution thermal measurements taken by the International Space Station over the Amazon, Congo Basin and southeast Asia. They found that during dry periods, canopy temperatures averaged around 34 °C, but that some can exceed 40 °C. Crucially, upper-canopy leaves exceeded the critical temperature 0.01% of the time; and in experiments that artificially warmed the upper canopies this rose to 1.3% of the time. Combined with ground-based observational and experimental data, the team estimates that large-scale leaf death and loss might begin to happen if warming exceeds 3.9 °C; the worst-case scenario for climate-change projections currently suggests air temperature would rise by 4 °C.