AI research

Summary OpenProteinSet provides a massive dataset of the same quality as the one used to train AlphaFold 2, which was not made available to the research community. Proteins are the workhorses of life. Understanding their sequences and structures is key to tackling challenges ranging from designing new enzymes to developing life-saving drugs. In recent years, …

OpenProteinSet provides open source training data for structural biology at scale Read More »

AI language models hold different political biases, according to a study from the University of Washington, Carnegie Mellon University, and Xi’an Jiaotong University. The researchers tested 14 models and found that OpenAI’s ChatGPT and GPT-4 were the most left-libertarian, while Meta’s LLaMA was the most right-wing authoritarian. The political compass test helped identify the models’ …

Researchers discover political biases within AI language models Read More »

Summary Companies like Google are developing language models optimized for medical purposes. Microsoft believes that GPT-4 is sufficient. According to Microsoft, large language models can help speed up medical processes by, for example, structuring “large unstructured data” that currently requires time-consuming manual processing. As an example, Microsoft cites the faster development of cancer drugs, where …

Microsoft Research says GPT-4 is good enough for medical tasks Read More »

Summary Dynalang is an AI agent that understands language and its environment by making predictions about the future in environments with a multimodal world model. A major challenge in AI research is to enable AI agents, such as robots, to communicate naturally with humans. Today’s agents, such as Google’s PaLM-SayCan, understand simple commands like “get …

AI agent Dynalang could change the way we talk to robots Read More »

Summary Various search engines for finding scientific publications are being equipped with AI functions. The companies promise to be especially careful, as AI-generated misinformation could do great harm, especially in science. On August 1, Elsevier, a large Dutch scientific publisher, released Scopus AI to a select group of 15,000 beta users. This is a GPT-3.5-based …

Scientific search engines cautiously test AI capabilities Read More »

Summary Under certain circumstances, AI models generalize beyond training data. This phenomenon is called “grokking” in AI research, and Google is now providing insight into recent findings. During training, AI models sometimes seem to suddenly “understand” a problem even though they have only memorized training data. In AI research, this phenomenon is called “grokking,” a …

When Stochastic Parrots build models Read More »

Summary There are many commercial and open-source generative text AIs. A benchmark developed specifically for testing assistance tasks now shows that GPT-4 stands out in this segment. “AgentBench” is a standardized test specifically designed to measure the assistance capabilities of large language models in “real-world pragmatic missions”. All tests are conducted in real-time interactive environments. …

GPT-4 dominates other LLMs in “real-world pragmatic missions,” study finds Read More »

Summary In a pilot project, Google and American Airlines are investigating whether planes can avoid condensation trails through clever route planning. They can, but at a cost. Using satellite imagery, weather and flight data, the project team used an AI system to predict which routes would cause contrails. American Airlines pilots then flew adjusted routes …

AI-controlled flight paths reduce contrails to help protect the climate Read More »

Summary A team of researchers presents the first high-quality, real-time rendering solution for radiation fields with training times in minutes and rendering at over 100 FPS. Researchers from Inria, the Max Planck Institute for Informatics, and the Université Côte d’Azur have introduced a new approach for real-time, high-quality rendering of radiance fields acquired from real …

New method unlocks the future of real-time, high-quality visuals Read More »

Summary Researchers at Meta have developed new photorealistic synthetic datasets using Unreal Engine that enable more controlled and robust evaluation and training of AI vision systems. Meta researchers have introduced a family of synthetic image datasets called PUG (Photorealistic Unreal Graphics) that aim to provide new capabilities for the evaluation and training of AI vision …

Meta’s PUG brings synthetic and real-world data closer using Unreal Engine Read More »