Powerful AI models, and more — this week’s best science graphics –

Home / Artificial Intelligence / Powerful AI models, and more — this week’s best science graphics –

Powerful AI models, and more — this week’s best science graphics –

Three charts from the world of research, selected by Nature editors.

Accelerating artificial intelligence

Commercial firms are rolling out ever-larger artificial-intelligence models — each costing millions of dollars to train and run. They hope to develop systems with capabilities greater than the large language models (LLMs) underlying chatbots such as ChatGPT. LLMs are giant networks of computing units (also called artificial neurons), arranged in layers. An LLM’s size is measured in how many parameters it has — the adjustable values that describe the strength of the connections between neurons.

The drive to bigger AI models: Scatter plot showing the number of parameters used to train AI models from 1950 to 2022.

Source: Adapted from Our World in Data, and from J. Sevilla et al. Preprint at https://arxiv.org/abs/2202.05924 (2022).

The cost of cancers

The total cost of cancer to the global economy will reach 25.2 trillion international dollars between 2020 and 2050, according to an analysis. Tracheal, bronchial and lung cancer will account for the largest share of the costs, followed by colon and rectal cancer, breast cancer, liver cancer and leukaemia. Estimating the cost of cancer — a leading cause of death worldwide — could push countries to invest in prevention and research.

Which cancers cost the most? Bar chart comparing the cost in international dollars of five most expensive cancer types.

Source: S. Chen et al. JAMA Oncol. https://doi.org/10.1001/jamaoncol.2022.7826 (2023)

Where flash floods strike

Flash floods are a growing threat in some of the world’s driest regions, where relatively small amounts of rain can trigger a deluge because dry soils repel water rather than allowing it to soak in. An analysis of emergency events shows that, since 2000, such regions experienced less than half (47%) of deadly flash floods globally, yet saw almost three-quarters (74%) of related deaths. The majority of these disasters occurred in low- and middle-income countries.

Global flash-flood disasters. Map showing dryland and wetland flash-floods across the decades plotted by severity.

Source: EM-DAT, CRED/UC Louvain (www.emdat.be); Analysis by J. Yin et al.

Leave a Reply

Your email address will not be published.