Chinese researchers yesterday unveiled a new EEG capture device that for the first time recorded synchronous electroencephalogram signals from more than 1,000 people across different regions. The team says it solved two key technical hurdles: miniaturizing hardware while preserving signal fidelity, and overcoming network latency to achieve millisecond-level time alignment across devices and locations. The capability will generate large-scale, high-quality EEG datasets for training neural foundat

2026-07-23

Chinese researchers yesterday unveiled a new EEG capture device that for the first time recorded synchronous electroencephalogram signals from more than 1,000 people across different regions. The team says it solved two key technical hurdles: miniaturizing hardware while preserving signal fidelity, and overcoming network latency to achieve millisecond-level time alignment across devices and locations. The capability will generate large-scale, high-quality EEG datasets for training neural foundation models and advancing commercial brain-computer interface (BCI) technology, expanding AI training inputs from text, images and video to direct neural signals.

其他消息
2026-07-23

瑞銀近期與至少19家AI企業召開了峯會討論了中國開源模型的影響和Token成本優化等議題,這些企業都是各自垂直領域的頭部企業,在這些議題上十分有話語權。在跟這些企業交流之後,瑞銀得出了幾個結論: 1、AI行業正在進入“多模型時代”。未來市場不會由少數閉源大模型主導,而是形成前沿模型、開源模型以及企業定製模型共存的生態。中國開源模型的崛起正在改變競爭邏輯。AI競爭不再只是追求最高模型性能,而是轉向“性能、成本、效率”的綜合競爭。 2、企業將根據任務複雜度,在高性能模型和低成本模型之間動態選擇,以降低Token消耗。這種能力將成爲AI應用和基礎設施公司的核心競爭力。 3、基於基礎模型微調的行業專用模型也將成爲重要方向,推動AI從通用能力向垂直場景深入。 4、AWS、Azure等雲廠商受益於多模型生態和持續增長的AI需求,影響偏中性向好。 5、GPU需求不會被削弱,AI工作負載將向更多推理任務擴展,整體利好英偉達等算力供應商。 6、相比之下,軟件行業可能面臨壓力,企業爲控制AI成本,可能減少部分AI功能使用,同時傳統軟件公司通過多模型兼容能力獲得收益的難度較高。

2026-07-23

Interfax reports ZELENSKIY said he discussed ideas to revive peace talks with a US special envoy.