<?xml version="1.0"?>
<rss version="2.0" xml:base="link" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:fb="http://www.facebook.com/2008/fbml" xmlns:foaf="http://xmlns.com/foaf/0.1/" xmlns:media="http://www.rssboard.org/media-rss" xmlns:og="http://ogp.me/ns#" xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#" xmlns:schema="http://schema.org/" xmlns:sioc="http://rdfs.org/sioc/ns#" xmlns:sioct="http://rdfs.org/sioc/types#" xmlns:skos="http://www.w3.org/2004/02/skos/core#" xmlns:xsd="http://www.w3.org/2001/XMLSchema">
  <channel>
    <title>He Kaiming - South China Morning Post</title>
    <link>https://www.scmp.com/rss/524439/feed</link>
    <description>The latest news and updates on He Kaiming, a prominent Chinese AI researcher, currently a Distinguished Scientist part-time at Google DeepMind and an Associate Professor at MIT. Formerly with Facebook AI Research and Microsoft Research Asia, he is celebrated for co-proposing Deep Residual Networks (ResNets), a pivotal contribution to deep learning now ubiquitous in AI models. His expertise spans computer vision and deep learning, with notable work on image segmentation (Mask R-CNN) and...</description>
    <language>en</language>
    <image>
      <url>https://assets.i-scmp.com/static/img/icons/scmp-meta-1200x630.png</url>
      <title>He Kaiming - South China Morning Post</title>
      <link>https://www.scmp.com</link>
    </image>
    <atom:link href="https://www.scmp.com/rss/524439/feed" rel="self" type="application/rss+xml"/>
    <item>
      <author>Eunice Xu</author>
      <dc:creator>Eunice Xu</dc:creator>
      <description>DeepSeek’s proposed “mHC” architecture could transform the training of large language models (LLMs) – the technology behind artificial intelligence chatbots – as developers look for ways to scale models without simply adding more computing power.
However, experts cautioned that while the approach could prove far-reaching, it might still prove difficult to put into practice.
In a technical paper released last week, co-authored by DeepSeek founder and CEO Liang Wenfeng, the company proposed...</description>
      <guid isPermaLink="true">https://www.scmp.com/tech/big-tech/article/3338763/deepseek-pitches-new-route-scale-ai-researchers-call-more-testing?utm_source=rss_feed</guid>
      <link>https://www.scmp.com/tech/big-tech/article/3338763/deepseek-pitches-new-route-scale-ai-researchers-call-more-testing?utm_source=rss_feed</link>
      <pubDate>Mon, 05 Jan 2026 11:00:10 +0000</pubDate>
      <title>DeepSeek pitches new route to scale AI, but researchers call for more testing</title>
      <enclosure length="3000" type="image/jpeg" url="https://cdn.i-scmp.com/sites/default/files/styles/1280x720/public/d8/images/canvas/2026/01/05/f201a590-d695-4f1f-b8b0-a60c77dd3015_4618fc8b.jpg?itok=x6kqD1N0&amp;v=1767601134"/>
      <media:content height="2001" medium="image" type="image/jpeg" url="https://cdn.i-scmp.com/sites/default/files/styles/1280x720/public/d8/images/canvas/2026/01/05/f201a590-d695-4f1f-b8b0-a60c77dd3015_4618fc8b.jpg?itok=x6kqD1N0&amp;v=1767601134" width="3000"/>
    </item>
    <item>
      <author>Eunice Xu</author>
      <dc:creator>Eunice Xu</dc:creator>
      <description>DeepSeek’s latest technical paper, co-authored by the firm’s founder and CEO Liang Wenfeng, has been cited as a potential game changer in developing artificial intelligence models, as it could translate into improvements in the fundamental architecture of machine learning.
The paper’s theme of Manifold-Constrained Hyper-Connections (mHC) marks an improvement to conventional hyper-connections and residual networks (ResNet), a fundamental mechanism underlying large language models (LLMs),...</description>
      <guid isPermaLink="true">https://www.scmp.com/tech/tech-trends/article/3338535/deepseek-proposes-shift-ai-model-development-mhc-architecture-upgrade-resnet?utm_source=rss_feed</guid>
      <link>https://www.scmp.com/tech/tech-trends/article/3338535/deepseek-proposes-shift-ai-model-development-mhc-architecture-upgrade-resnet?utm_source=rss_feed</link>
      <pubDate>Fri, 02 Jan 2026 10:30:15 +0000</pubDate>
      <title>DeepSeek proposes shift in AI model development with ‘mHC’ architecture to upgrade ResNet</title>
      <enclosure length="3000" type="image/jpeg" url="https://cdn.i-scmp.com/sites/default/files/styles/1280x720/public/d8/images/canvas/2026/01/02/d45aba8e-e7c4-41a9-809b-e8fcd9bc6055_1f789d0d.jpg?itok=ZJm4TRCM&amp;v=1767347421"/>
      <media:content height="2001" medium="image" type="image/jpeg" url="https://cdn.i-scmp.com/sites/default/files/styles/1280x720/public/d8/images/canvas/2026/01/02/d45aba8e-e7c4-41a9-809b-e8fcd9bc6055_1f789d0d.jpg?itok=ZJm4TRCM&amp;v=1767347421" width="3000"/>
    </item>
    <item>
      <author>Vincent Chow</author>
      <dc:creator>Vincent Chow</dc:creator>
      <description>Chinese artificial intelligence start-up DeepSeek has ushered in 2026 with a new technical paper, co-authored by founder Liang Wenfeng, that proposes a rethink of the fundamental architecture used to train foundational AI models.
The method – dubbed Manifold-Constrained Hyper-Connections (mHC) – forms part of the Hangzhou firm’s push to make its models more cost-effective as it strives to keep pace with better-funded US rivals with deeper access to computing power.
It also reflected the...</description>
      <guid isPermaLink="true">https://www.scmp.com/tech/big-tech/article/3338427/deepseek-kicks-2026-paper-signalling-push-train-bigger-models-less?utm_source=rss_feed</guid>
      <link>https://www.scmp.com/tech/big-tech/article/3338427/deepseek-kicks-2026-paper-signalling-push-train-bigger-models-less?utm_source=rss_feed</link>
      <pubDate>Thu, 01 Jan 2026 13:51:48 +0000</pubDate>
      <title>DeepSeek kicks off 2026 with paper signalling push to train bigger models for less</title>
      <enclosure length="4095" type="image/jpeg" url="https://cdn.i-scmp.com/sites/default/files/styles/1280x720/public/d8/images/canvas/2026/01/01/8280bbbd-fa54-4df2-8a69-28f069c3bd18_9e1188c1.jpg?itok=N7v162jA&amp;v=1767273963"/>
      <media:content height="3092" medium="image" type="image/jpeg" url="https://cdn.i-scmp.com/sites/default/files/styles/1280x720/public/d8/images/canvas/2026/01/01/8280bbbd-fa54-4df2-8a69-28f069c3bd18_9e1188c1.jpg?itok=N7v162jA&amp;v=1767273963" width="4095"/>
    </item>
    <item>
      <author>Ben Jiang</author>
      <dc:creator>Ben Jiang</dc:creator>
      <description>Facebook parent Meta Platforms is adding another Chinese artificial intelligence (AI) expert to its newly established research facility, underscoring how researchers from China are increasingly sought after in the competitive global race for top tech talent.
The social media giant has recruited Pang Ruoming, who held the title of distinguished software engineer at Apple and was leading its foundation models team, according to a Bloomberg News report on Tuesday that was confirmed by Meta.
Neither...</description>
      <guid isPermaLink="true">https://www.scmp.com/tech/big-tech/article/3317370/metas-ai-lab-scoops-more-chinese-experts-apple-openai-aggressive-talent-grab?utm_source=rss_feed</guid>
      <link>https://www.scmp.com/tech/big-tech/article/3317370/metas-ai-lab-scoops-more-chinese-experts-apple-openai-aggressive-talent-grab?utm_source=rss_feed</link>
      <pubDate>Tue, 08 Jul 2025 10:30:07 +0000</pubDate>
      <title>Meta’s AI lab scoops up more Chinese experts from Apple, OpenAI in aggressive talent grab</title>
      <enclosure length="4095" type="image/jpeg" url="https://cdn.i-scmp.com/sites/default/files/styles/1280x720/public/d8/images/canvas/2025/07/08/d2c9d741-e4cc-4dff-b096-4d6f55d63926_6f2ff97b.jpg?itok=ZPU-qwXZ&amp;v=1751961217"/>
      <media:content height="2730" medium="image" type="image/jpeg" url="https://cdn.i-scmp.com/sites/default/files/styles/1280x720/public/d8/images/canvas/2025/07/08/d2c9d741-e4cc-4dff-b096-4d6f55d63926_6f2ff97b.jpg?itok=ZPU-qwXZ&amp;v=1751961217" width="4095"/>
    </item>
  </channel>
</rss>