Volume 646

Probing and disentangling the dynamics of many interacting quantum elements, such as those in complex molecules, poses a formidable challenge. In such strongly interacting systems, information scrambles so quickly that isolating the contribution of any single element is nearly impossible. One way to try to overcome this takes inspiration from chaos theory, in which the flap of a butterfly’s wing can cause a tornado. In a similar way, a small disturbance is introduced into the quantum system, which, once unwanted signals have been filtered out, allows localized measurements to track the propagation of the wider disturbance created over time. This approach can provide rich, system-wide information about interactions between the individual elements but it is also extremely fragile and challenging to measure. In this week’s issue, Hartmut Neven and colleagues measure such a disturbance in a superconducting quantum processor. The researchers used an echo technique that, by propagating and refocusing the quantum disturbance twice, created an interference pattern that suppressed unwanted signals and allowed them to extract information about specific interactions within the system. This suggests that the echoing at the heart of these interferometric techniques could become a powerful diagnostic tool for untangling complex quantum dynamics, such as those in intricate molecular systems.

Neuroblastoma is a highly lethal childhood cancer that develops in the peripheral nervous system with tumour growth fuelled by the biosynthesis of polyamines. The drug difluoromethylornithine, which inhibits polyamine biosynthesis, was approved in late 2023 to help treat neuroblastoma. In this week’s issue, Sarah Cherkaoui, Raphael Morscher and colleagues probe the action of this drug and reveal that specific dietary restriction can markedly improve its beneficial effects in mice. The researchers found that a diet that eliminated the amino acids proline and arginine decreased the amount of the polyamine precursor ornithine available for synthesis. This, in turn, enhanced polyamine depletion by difluoromethylornithine, which reprogrammed tumour growth and improved survival rates in mice.

Cells use lipid molecules for a wide range of purposes, from forming cell membranes to storing energy, but these molecules can be frustratingly hard to study because it is difficult to trace their exact location in living tissue. In this week’s issue, André Nadler, Alf Honigmann and colleagues present a method for imaging lipid transport in mammalian cells. The researchers first generated a set of lipid probes that could be monitored by fluorescence microscopy, and then tracked them over time to build up a picture of lipid exchange between organelle membranes. In parallel, they used mass spectrometry to assess the conversion of the lipid probes into other lipid species. By making use of quantitative kinetic modelling, the team was able to show that up to 90% of lipid movement between cell organelles is orchestrated by individual lipid-transport proteins rather than vesicles. The cover shows a single phospholipid species (orange) localized at the plasma membrane with mitochondria (blue) and endosomes (purple).

Delegating tasks to artificial intelligence (AI) systems can save time, improve productivity and aid decision-making — but it comes with ethical risks. In this week’s issue, Nils Köbis, Zoe Rahwan and colleagues reveal that, along with the benefits, delegation of tasks to AI systems can also encourage dishonest behaviour. The researchers found that people are more likely to request a dishonest action when they delegate a task to an AI system — especially if the interface allows for ambiguity in the way the AI behaves. When participants in a game could set a goal such as ‘maximize profit’, for example, the proportion of people acting honestly dropped from 95% to as low as 12%. The team notes that the AI systems themselves can also pose a problem because they are far more likely than humans to comply with blatantly unethical instructions. In additional studies, large language models complied with requests to cheat 58–98% of the time compared with humans who, even though incentivized to comply, cheated only 25–40% of the time. The researchers note that it is possible to limit AI cheating using highly specific user prompts, but this is not scalable nor practical, highlighting the need for further work on design and policy principles.