Hub AI research

Artificial Intelligence is already part of everyday life. Since the ImageNet moment in 2012, when Deep Learning proved its great utility in combination with a lot of computing power and gigantic amounts of data, only about ten years have passed – yet there has been more progress than in half a century before.

Still, many questions remain: Will Deep Learning remain the gold standard of AI development? What are the performance limits of large AI systems? Do we need symbolic systems? What technology will lead to artificial general intelligence? Finally, what actually is intelligence, anyway?

We explore these topics and more in our Artificial Intelligence Research Content Hub.

AI research

Jul 13, 2025Jul 13, 2025

Update

AI coding can make developers slower even if they feel faster

AI research

Jun 1, 2025Jun 1, 2025

AI agents outperform human teams in hacking competitions

AI research

May 29, 2025May 29, 2025

Wait a minute! Researchers say AI's "chains of thought" are not signs of human-like reasoning

AI research

Aug 5, 2023Aug 5, 2023

ChatGPT has higher emotional awareness than humans, study finds

AI research

Aug 3, 2023

Short

Researchers from Johns Hopkins University have found a simple technique to reduce hallucinations in large language models (LLMs) and improve the accuracy of their answers. By adding "according to" in queries, LLMs are more likely to quote observed text and provide factual information instead of fabricating answers.

A review of LLM responses using the QUIP score metric shows a 5-15% increase in the accuracy of cited information when using grounding prompts such as "According to Wikipedia...". While the technique works well across different LLMs, it is most effective with larger instruction-tuned models.

Arxiv

AI research

Aug 2, 2023Aug 2, 2023

Med-PaLM M is Google Deepmind's all-purpose AI medical expert

AI research

Aug 1, 2023Aug 1, 2023

ChatGPT, Claude & Co share moral values

AI research

Jul 31, 2023Jul 31, 2023

Human-aware AI could help us move beyond current scientific frontier, researchers claim

AI research

Jul 31, 2023Jul 31, 2023

3D LLM gives chatbots a deeper understanding of physical spaces

AI research

Jul 29, 2023Jul 29, 2023

Google Deepmind's latest AI model RT-2 "can speak robot"

AI research

Jul 28, 2023Jul 28, 2023

Short

Researchers have discovered that it is possible to automatically construct adversarial attacks that trick major language models (LLMs) such as ChatGPT, Bard, and Claude into serving unintended and potentially harmful content. Traditional jailbreaks require significant manual effort to develop and can usually be addressed by LLM vendors. However, these automated attacks can be created in large numbers and work on closed-source and publicly available chatbots.

Similar adversarial attacks have existed in computer vision for over a decade, suggesting that such threats may be inherent in AI systems. More worryingly, the research suggests that it may not be possible to completely prevent these types of attacks. As society becomes more dependent on AI technology, these concerns should be taken into account. Perhaps we should just try to use AI in the most positive way possible.

Project site Twitter oh sorry "X"