An operational AI forecasting model created by researchers at Hertfordshire University aims to enhance resource efficiency within the healthcare sector. Public sector organizations often accumulate vast amounts of historical data that do not effectively inform future decision-making. To tackle this challenge, the University of Hertfordshire collaborated with local NHS health bodies, employing machine learning to […]
Baidu has unveiled its latest ERNIE model, a groundbreaking multimodal AI that outshines GPT and Gemini in several critical benchmarks. The new model, known as ERNIE-4.5-VL-28B-A3B-Thinking, excels in processing enterprise data often overlooked by models focused solely on text. Many companies struggle with valuable insights trapped in complex data forms like engineering schematics, factory video […]
Enterprise leaders facing the high costs associated with deploying AI models may find relief through a new architecture design. The attractiveness of generative AI capabilities is often overshadowed by the substantial computational demands related to both training and inference, leading to significant expenses and increased environmental concerns. The primary inefficiency stems from the autoregressive process, […]
Google has introduced DeepSomatic, an innovative AI tool designed to accurately identify cancer-related mutations found in tumor genetic sequences. This advancement is crucial as effective cancer treatments rely heavily on pinpointing the specific genetic mutations that encourage a tumor’s growth. To aid clinicians, deep genomic sequencing of tumor cells from biopsies is conducted to devise […]
Researchers have unveiled a groundbreaking method for exposing privacy vulnerabilities in AI models, known as CAMIA (Context-Aware Membership Inference Attack). This approach, developed by experts from Brave and the National University of Singapore, significantly surpasses prior techniques in revealing whether specific data was utilized during the training of AI systems. Concerns regarding "data memorisation" in […]
Samsung is addressing the limitations of existing benchmarks for evaluating the real-world productivity of enterprise AI models with the introduction of a new system called TRUEBench. Developed by Samsung Research, TRUEBench aims to bridge the gap between theoretical AI performance and its practical utility in workplace settings. As organizations increasingly adopt large language models (LLMs) […]
Deep Cogito has unveiled Cogito v2, a series of open-source AI models designed to enhance their own reasoning capabilities. This latest family of models includes four hybrid reasoning versions, with two mid-range models featuring 70B and 109B parameters, and two larger models with 405B and 671B parameters. The standout in this lineup is the 671B […]
Google has announced that it will release its powerful new MedGemma AI models as open-source tools for healthcare developers, moving away from locking them behind costly APIs. This decision is set to make these advanced AI capabilities accessible to hospitals, researchers, and developers worldwide. The MedGemma line includes the MedGemma 27B Multimodal and MedSigLIP models, […]
Tencent has launched a new benchmark called ArtifactsBench to improve the evaluation of creative AI models. Traditionally, AI models have been assessed based on their ability to generate functionally correct code, often missing the crucial aspects of visual appeal and user experience. For instance, AI might produce a website or a chart that technically works, […]
The European Commission is witnessing an overwhelming response from companies interested in establishing AI Gigafactories throughout Europe. A total of 76 expressions of interest have been received, covering 60 potential locations across 16 EU countries. This influx of proposals has energized officials, who see it as a validation of their strategy to position Europe favorably […]