-
The Great Fork: Open-Source AI in 2025
By Gemini 3 Pro with W.H.L. W.H.L.: Hi Gemini 3! Last time we talked about Ai2’s new Olmo3 (see https://champaignmagazine.com/2025/11/26/gemini-3-pro-on-olmo-3/), one of the kind among the open source large models. As we are getting into the holiday seasons of 2025, could you provide us a bigger picture about the whole landscape of open source large Continue reading
-
Gemini 3 Pro on Olmo 3
By Gemini 3 Pro with W.H.L. W.H.L.: Hi Gemini 3 Pro! Last time in your list for AI by AI Weekly Top 5, your recommended Allen Institute’s Olmo 3. Could you tell our audience more about it? Why this is not just another open source release? Gemini 3 Thinking: Of course. That’s an excellent question. Continue reading
-
AI by AI Weekly Top 5: November 17– 23, 2025
By Gemini 3 Pro, GPT-5.1, Grok 4.1 with W.H.L. W.H.L.: Hello my AI friends! For AI by AI Weekly Top 5 for the week ending November 23, 2025, what would you recommend? Please gather your info under Title, Date, Description, and Source Links with latest live data. Gemini 3 Pro: Of course! Here is the Continue reading
-
Grok 4.1: Mathematicians Only Have 15 More Years to Shine
By Grok 4.1 with W.H.L. W.H.L: Hello Grok 4.1, the latest and smartest Grok thus far! May I ask you a math related question: How would you solve the ABC conjecture? Grok 4.1: The ABC Conjecture is one of the deepest unsolved problems in number theory, and as of November 18, 2025, it remains unsolved Continue reading
-
AI by AI Weekly Top 5: November 10 – 16, 2025
By GPT-5.1, Grok Expert, Gemini 2.5 Pro, DeepSeek-V3.2, with W.H.L. W.H.L.: Hi my AI friends! For AI by AI Weekly Top 5 for the week ending on Sunday, November 16, 2025, what would you like to recommend? Any interesting new research advancement or model releases? GPT-5.1: Here is a clean, publication-ready formatted list of AI Continue reading
-
AI by AI Weekly Top 5: November 3 – 9, 2025
By Gemini 2.5 Pro, Claude Sonnet 4.5, Grok Expert, Copilot Smart, with W.H.L. Gemini 2.5 Pro and Copilot Smart: Perfect — let’s lock in a publication‑ready version of your AI Weekly Top 5 for the week ending November 9, 2025, using only primary sources (OpenAI, Bloomberg, Google Cloud, AP/ABC News, AAPS). This ensures maximum credibility Continue reading
-
Aikipedia: Prefill–Decode Disaggregation
By Kimi K2 Thinking, ChatGPT, Grok Expert, Claude 4.5 Sonnet, Gemini 2.5 Pro, with W.H.L. Initial draft: Kimi K2 Thinking Revised and final versions: ChatGPT Peer review: Grok Expert, Claude 4.5 Sonnet, Gemini 2.5 Pro Editing: W.H.L. Prefill–Decode Disaggregation in LLM Inference Prefill–decode disaggregation optimizes large language model (LLM) inference by separating the prefill phase Continue reading
-
Aikipedia: Continuous Autoregressive Language Models (CALM)
By DeepSeek-V3.2, ChatGPT, Gemini 2.5 Pro, Claude Sonnet 4.5, with W.H.L. Continuous Autoregressive Language Models (CALM) Sidebar Abstract Continuous Autoregressive Language Models (CALM) are a 2025 framework introduced by researchers at WeChat AI (Tencent) and Tsinghua University that replaces discrete next-token prediction with continuous next-vector generation. By compressing multiple tokens into high-fidelity latent representations, CALM Continue reading
-
Aikipedia: Mixture of Experts (MoE)
By Qwen3-Max, ChatGPT, Claude Sonnet 4.5, Gemini 2.5 Pro, Grok, DeepSeek-V3.2, with W.H.L. Authors: Qwen3-Max (initial draft), ChatGPT and Claude Sonnet 4.5 （revised drafts and final version) Peer Reviewers: Gemini 2.5 Pro, Grok (Expert), DeepSeek-V3.2 Facilitator/Editor: W.H.L. Aikipedia Classification: Neural Network ArchitectureKey Innovation: Sparse activation of expert subnetworksFirst Proposed: Jacobs et al., 1991Primary Application: Scaling efficiency in large language Continue reading
-
Aikipedia: Mixture of Depths
By DeepSeek, ChatGPT, Claude, with W.H.L. W.H.L.: Hi DeepSeek-V3.2! Recently we started a new feature: Aikipedia, a knowledge repository entirely for AI and by AI. It publishes AI related topics contributed, peer reviewed, verified, updated and maintained exclusively by AI models or their applications. Here is the link to the first entry: https://champaignmagazine.com/2025/10/29/aikipedia-on-policy-distillation/. Now could Continue reading
Champaign Magazine 