Featured
-
-
Article
| Open AccessLarger and more instructable language models become less reliable
Scaling up and shaping up large language models increased their tendency to provide sensible yet incorrect answers at difficulty levels humans cannot supervise, highlighting the need for a fundamental shift in artificial intelligence design towards reliability.
- Lexin Zhou
- , Wout Schellaert
- & José Hernández-Orallo
-
Article
| Open AccessBendable non-silicon RISC-V microprocessor
Flex-RV, a 32-bit microprocessor based on an open RISC-V instruction set fabricated with indium gallium zinc oxide thin-film transistors on a flexible polyimide substrate, enables an ultralow-cost bendable and flexible microprocessor.
- Emre Ozer
- , Jedrzej Kufel
- & Vijay Janapa Reddi
-
Editorial |
How to stop a looming ‘splinternet’
Online safety is crucial, but so are privacy and decentralization. Computer scientists who set the Internet’s technical standards should be included in governance talks.
-
News |
Do AI models produce more original ideas than researchers?
The concepts were judged by reviewers. They were not told who or what had created them.
- Gemma Conroy
-
Technology Feature |
Forget ChatGPT: why researchers now run small AIs on their laptops
Artificial-intelligence models are typically used online, but a host of openly available tools is changing that. Here’s how to get started with local AIs.
- Matthew Hutson
-
Career Guide |
Guide, don’t hide: reprogramming learning in the wake of AI
As artificial intelligence becomes increasingly integral to the world outside academia, universities face a crucial choice: to use or not to use.
- Monique Brouillette
-
News Feature |
A day in the life of the world’s fastest supercomputer
In the hills of eastern Tennessee, a record-breaking machine called Frontier is providing scientists with unprecedented opportunities to study everything from atoms to galaxies.
- Sophia Chen
-
Research Briefing |
Holistic approach to carbon capture bridges the ‘Valley of Death’
Carbon-capture technology often founders at the point when basic research is translated into practical applications. A computational modelling platform called PrISMa solves this problem by considering the needs of all stakeholders.
-
Article
| Open AccessAI generates covertly racist decisions about people based on their dialect
Despite efforts to remove overt racial prejudice, language models using artificial intelligence still show covert racism against speakers of African American English that is triggered by features of the dialect.
- Valentin Hofmann
- , Pratyusha Ria Kalluri
- & Sharese King
-
News |
Science treasures from Microsoft mogul up for auction — and researchers are salivating
Spacesuits, historic computers and more from the estate of the late Paul Allen are going on sale.
- Alix Soliman
-
News |
AI made of jelly ‘learns’ to play Pong — and improves with practice
Inspired by neurons in a dish playing the classic video game, researchers show that synthetic hydrogels have a basic ‘memory’.
- Gemma Conroy
-
Comment |
Light bulbs have energy ratings — so why can’t AI chatbots?
The rising energy and environmental cost of the artificial-intelligence boom is fuelling concern. Green policy mechanisms that already exist offer a path towards a solution.
- Sasha Luccioni
- , Boris Gamazaychikov
- & Carole-Jean Wu
-
News & Views |
Switching between tasks can cause AI to lose the ability to learn
Artificial neural networks become incapable of mastering new skills when they learn them one after the other. Researchers have only scratched the surface of why this phenomenon occurs — and how it can be fixed.
- Clare Lyle
- & Razvan Pascanu
-
Article
| Open AccessLoss of plasticity in deep continual learning
The pervasive problem of artificial neural networks losing plasticity in continual-learning settings is demonstrated and a simple solution called the continual backpropagation algorithm is described to prevent this issue.
- Shibhansh Dohare
- , J. Fernando Hernandez-Garcia
- & Richard S. Sutton
-
Nature Video |
Why ChatGPT can't handle some languages
In a test of the chatbot's language abilities it fails at certain languages.
- Nick Petrić Howe
-
Nature Podcast |
ChatGPT has a language problem — but science can fix it
The Large Language Models that power chatbots are known to struggle in languages outside of English — this podcast explores how this challenge can be overcome.
- Nick Petrić Howe
-
Article
| Open AccessFully forward mode training for optical neural networks
We present fully forward mode learning, which conducts machine learning operations on site, leading to faster learning and promoting advancement in numerous fields.
- Zhiwei Xue
- , Tiankuang Zhou
- & Lu Fang
-
Technology Feature |
Quantum computing aims for diversity, one qubit at a time
The fast-growing discipline needs more scientists from under-represented groups. A raft of initiatives is rising to the challenge.
- Amanda Heidt
-
Career Feature |
Slow productivity worked for Marie Curie — here’s why you should adopt it, too
Do fewer things, work at a natural pace and obsess over quality, says computer scientist Cal Newport, in his latest time-management book.
- Anne Gulland
-
Outlook |
AI is vulnerable to attack. Can it ever be used safely?
The models that underpin artificial-intelligence systems such as ChatGPT can be subject to attacks that elicit harmful behaviour. Making them safe will not be easy.
- Simon Makin
-
News & Views |
AI produces gibberish when trained on too much AI-generated data
Generative AI models are now widely accessible, enabling everyone to create their own machine-made something. But these models can collapse if their training data sets contain too much AI-generated content.
- Emily Wenger
-
Article
| Open AccessAI models collapse when trained on recursively generated data
Analysis shows that indiscriminately training generative artificial intelligence on real and generated content, usually done by scraping data from the Internet, can lead to a collapse in the ability of the models to generate diverse high-quality output.
- Ilia Shumailov
- , Zakhar Shumaylov
- & Yarin Gal
-
Technology Feature |
ChatGPT for science: how to talk to your data
Companies are using artificial intelligence tools to help scientists to query their data without the need for programming skills.
- Julian Nowogrodzki
-
News |
Can AI be superhuman? Flaws in top gaming bot cast doubt
Building robust AI systems that always outperform people might be harder than thought, say researchers who studied Go-playing bots.
- Matthew Hutson
-
Technology Feature |
Inside the maths that drives AI
Loss functions measure algorithmic errors in artificial-intelligence models, but there’s more than one way to do that. Here’s why the right function is so important.
- Michael Brooks
-
World View |
How I’m using AI tools to help universities maximize research impacts
Artificial-intelligence algorithms could identify scientists who need support with translating their work into real-world applications and more. Leaders must step up.
- Dashun Wang
-
News & Views |
‘Fighting fire with fire’ — using LLMs to combat LLM hallucinations
The number of errors produced by an LLM can be reduced by grouping its outputs into semantically similar clusters. Remarkably, this task can be performed by a second LLM, and the method’s efficacy can be evaluated by a third.
- Karin Verspoor
-
Article
| Open AccessDetecting hallucinations in large language models using semantic entropy
Hallucinations (confabulations) in large language model systems can be tackled by measuring uncertainty about the meanings of generated responses rather than the text itself to improve question-answering accuracy.
- Sebastian Farquhar
- , Jannik Kossen
- & Yarin Gal
-
Article |
Experiment-free exoskeleton assistance via learning in simulation
A learning-in-simulation framework for wearable robots uses dynamics-aware musculoskeletal and exoskeleton models and data-driven reinforcement learning to bridge the gap between simulation and reality without human experiments to assist versatile activities.
- Shuzhen Luo
- , Menghan Jiang
- & Hao Su
-
News & Views |
Meta’s AI translation model embraces overlooked languages
More than 7,000 languages are in use throughout the world, but popular translation tools cannot deal with most of them. A translation model that was tested on under-represented languages takes a key step towards a solution.
- David I. Adelani
-
Article
| Open AccessScaling neural machine translation to 200 languages
Scaling neural machine translation to 200 languages is achieved by No Language Left Behind, a single massively multilingual model that leverages transfer learning across languages.
- Marta R. Costa-jussà
- , James Cross
- & Jeff Wang
-
News Feature |
How cutting-edge computer chips are speeding up the AI revolution
Engineers are harnessing the powers of graphics processing units (GPUs) and more, with a bevy of tricks to meet the computational demands of artificial intelligence.
- Dan Garisto
-
News Explainer |
Who owns your voice? Scarlett Johansson OpenAI complaint raises questions
In the age of artificial intelligence, situations are emerging that challenge the laws over rights to a persona.
- Nicola Jones
-
Article
| Open AccessLow-latency automotive vision with event cameras
Use of a 20 frames per second (fps) RGB camera plus an event camera can achieve the same latency as a 5,000-fps camera with the bandwidth of a 45-fps camera without compromising accuracy.
- Daniel Gehrig
- & Davide Scaramuzza
-
Correspondence |
Anglo-American bias could make generative AI an invisible intellectual cage
- Queenie Luo
- & Michael Puett
-
Editorial |
AlphaFold3 — why did Nature publish it without its code?
Criticism of our decision to publish AlphaFold3 raises important questions. We welcome readers’ views.
-
News |
AI now beats humans at basic tasks — new benchmarks are needed, says major report
Stanford University’s 2024 AI Index charts the meteoric rise of artificial-intelligence tools.
- Nicola Jones
-
Article
| Open AccessHigh-threshold and low-overhead fault-tolerant quantum memory
An end-to-end quantum error correction protocol that implements fault-tolerant memory on the basis of a family of low-density parity-check codes shows the possibility of low-overhead fault-tolerant quantum memory within the reach of near-term quantum processors.
- Sergey Bravyi
- , Andrew W. Cross
- & Theodore J. Yoder
-
Nature Podcast |
AI hears hidden X factor in zebra finch love songs
Machine learning detects song differences too subtle for humans to hear, and physicists harness the computing power of the strange skyrmion.
- Nick Petrić Howe
- & Benjamin Thompson
-
Correspondence |
Three reasons why AI doesn’t model human language
- Johan J. Bolhuis
- , Stephen Crain
- & Andrea Moro
-
Technology Feature |
So … you’ve been hacked
Research institutions are under siege from cybercriminals and other digital assailants. How do you make sure you don’t let them in?
- Michael Brooks
-
Technology Feature |
No installation required: how WebAssembly is changing scientific computing
Enabling code execution in the web browser, the multilanguage tool is powerful but complicated.
- Jeffrey M. Perkel
-
Editorial |
Why scientists trust AI too much — and what to do about it
Some researchers see superhuman qualities in artificial intelligence. All scientists need to be alert to the risks this creates.
-
News Explainer |
Is ChatGPT making scientists hyper-productive? The highs and lows of using AI
Large language models are transforming scientific writing and publishing. But the productivity boost that these tools bring could have a downside.
- McKenzie Prillaman
-
World View |
Generative AI’s environmental costs are soaring — and mostly secret
First-of-its-kind US bill would address the environmental costs of the technology, but there’s a long way to go.
- Kate Crawford
-
Editorial |
Cyberattacks on knowledge institutions are increasing: what can be done?
For months, ransomware attacks have debilitated research at the British Library in London and Berlin’s natural history museum. They show how vulnerable scientific and educational institutions are to this kind of crime.
-
News |
AI chatbot shows surprising talent for predicting chemical properties and reactions
Researchers lightly tweak ChatGPT-like system to offer chemistry insight.
- Davide Castelvecchi
-
Editorial |
How can scientists make the most of the public’s trust in them?
Researchers have a part to play in addressing concerns about government interference in science.
-