OpenAI’s math gold hints that AI may soon tackle even longer and harder tasks
OpenAI’s math gold hints that AI may soon tackle even longer and harder tasks An unreleased OpenAI model has reportedly solved five out of six International Mathematical Olympiad (IMO) problems under competition conditions. But the story isn’t just about what it solved – it’s about how it got there. The article OpenAI’s math gold hints […]
AI training shifts from clickworkers to experts in physics, biology and engineering
AI training shifts from clickworkers to experts in physics, biology and engineering Instead of relying on low-cost clickworkers, companies like Scale AI, Toloka, and Turing are now turning to highly skilled experts in fields like physics, biology, software engineering, and finance. The article AI training shifts from clickworkers to experts in physics, biology and engineering […]
New ARC-AGI-3 benchmark shows that humans still outperform LLMs at pretty basic thinking
New ARC-AGI-3 benchmark shows that humans still outperform LLMs at pretty basic thinking ARC-AGI-3 aims to test how well AI systems can handle brand new problems. While people breeze through the challenges, the latest AI models still come up short. The article New ARC-AGI-3 benchmark shows that humans still outperform LLMs at pretty basic thinking […]
Alibaba’s Qwen2.5 only excels at math thanks to memorized training data
Alibaba’s Qwen2.5 only excels at math thanks to memorized training data A new study finds that Alibaba’s Qwen2.5 models achieve high math scores mainly by memorizing training data rather than through genuine reasoning. The article Alibaba’s Qwen2.5 only excels at math thanks to memorized training data appeared first on THE DECODER. Read More
OpenAI claims a breakthrough in LLM reasoning on complex math problems
OpenAI claims a breakthrough in LLM reasoning on complex math problems OpenAI says its experimental language model has solved International Mathematical Olympiad (IMO) problems at a gold medal level—a possible breakthrough for AI with general reasoning skills. The results have not yet been independently confirmed. The article OpenAI claims a breakthrough in LLM reasoning on […]
Nvidia opens CUDA platform for RISC-V processors
Nvidia opens CUDA platform for RISC-V processors Nvidia announced at the RISC-V Summit China that it will open up its CUDA platform to support RISC-V processors. The article Nvidia opens CUDA platform for RISC-V processors appeared first on THE DECODER. Read More
Netflix used generative AI to create a VFX scene in its series “El Eternauta”
Netflix used generative AI to create a VFX scene in its series “El Eternauta” Netflix used generative AI to produce the final VFX scene in its Argentinian series “El Eternauta.” The article Netflix used generative AI to create a VFX scene in its series “El Eternauta” appeared first on THE DECODER. Read More
Yet another study finds that overloading LLMs with information leads to worse results
Yet another study finds that overloading LLMs with information leads to worse results Large language models are supposed to handle millions of tokens – the fragments of words and characters that make up their inputs – at once. But the longer the context, the worse their performance gets. The article Yet another study finds that […]
Google DeepMind’s Gemini wins Mathematical Olympiad gold using only natural language
Google DeepMind’s Gemini wins Mathematical Olympiad gold using only natural language Google DeepMind has claimed its first gold medal at the International Mathematical Olympiad (IMO) with an “advanced version” of its Gemini model running in Deep Think mode. The article Google DeepMind’s Gemini wins Mathematical Olympiad gold using only natural language appeared first on THE […]
FlexOlmo enables organizations to collaboratively train LLMs without data sharing
FlexOlmo enables organizations to collaboratively train LLMs without data sharing FlexOlmo, developed by the Allen Institute for AI, shows that organizations can collaboratively train language models on local datasets without sharing sensitive data. The article FlexOlmo enables organizations to collaboratively train LLMs without data sharing appeared first on THE DECODER. Read More