Champaign Magazine

champaignmagazine.com

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 developments strictly from the week of Nov 10–16, 2025 — combining your Top 5 plus the additional items that week. I’ve grouped them and kept each entry tight, with dates + sources.

AI by AI Weekly Top 5 — Week Ending Nov 16, 2025

1. OpenAI Launches GPT-5.1

Date: Nov 12, 2025
Sources: OpenAI blog; MacRumors; Dataconomy
Summary: OpenAI rolled out GPT-5.1, a major upgrade featuring “Instant” and “Thinking” modes, improved reasoning, faster responses, and customizable conversational “vibes.”
Why it matters: The most consequential model update of the week — strengthens agentic behavior and personalization.

2. Cancer AI Alliance Announces 8 New AI-Driven Cancer Research Projects

Date: Nov 10–11, 2025
Source: Johns Hopkins Medicine Newsroom
Summary: CAIA launched new projects using AI models for rare cancer detection, treatment planning, and molecular outcome prediction.
Why it matters: A direct example of AI transforming medical research, aligning with domain-expert AI expansion.

3. NVIDIA’s Earth-2 Accelerates Global Weather Prediction with CorrDiff

Date: Nov 10, 2025
Source: NVIDIA Developer Blog
Summary: NVIDIA published results showing 50× faster high-resolution forecasting using the CorrDiff generative super-resolution model on Earth-2.
Why it matters: Real-world, high-impact application of AI simulation — key for climate, disaster response, energy planning.

4. XAIxArts Workshop Proceedings Released (ArXiv)

Date: Nov 13, 2025
Source: ArXiv
Summary: Proceedings highlight new research at the intersection of explainability, multimodality, and creative AI — including interpretable generative art systems.
Why it matters: Expands interpretability into creative domains; contributes to understanding human–AI collaboration.

5. Yann LeCun Reportedly Considering Departure from Meta

Date: Nov 12, 2025
Source: Le Monde
Summary: Reports emerged that AI pioneer Yann LeCun may leave Meta to launch an independent lab, amid internal strategic disagreements.
Why it matters: A potentially seismic leadership change in global AI research direction.

Additional Notable AI Events (Nov 10–16, 2025)

6. Meta Releases “Omnilingual ASR” for 1,600+ Languages

Date: Nov 10
Source: VentureBeat
Why it matters: Major leap in multilingual modeling; directly relevant to global AI accessibility.

7. OpenAI Reports Sparse Architectures Improve Model Debugging

Date: Nov 14
Source: VentureBeat
Why it matters: Crucial interpretability advances; supports controllability and safe deployment.

8. Study Finds LLMs Have Distinct “Memory” vs “Reasoning” Neural Regions

Date: Nov 10
Source: Ars Technica
Why it matters: Foundational scientific insight into how large models internalize information.

9. Gartner Predicts AI Will Touch All IT Work by 2030

Date: Nov 10
Source: Gartner
Why it matters: Captures the macro-trend of widespread AI integration across industries.

W.H.L.: Grok, could you fact check, date check and source verify this list?

Grok Expert:

OpenAI Launches GPT-5.1
Verification: Fully confirmed. Launch on Nov 12, 2025; details match (Instant/Thinking modes, reasoning upgrades); sources align (OpenAI blog, MacRumors, etc.).

Cancer AI Alliance Announces 8 New AI-Driven Cancer Research Projects
Verification: Partially confirmed with correction. Announced Nov 10–11, 2025; only 2 projects (not 8), focused on detection/treatment; source (Johns Hopkins) matches.

NVIDIA’s Earth-2 Accelerates Global Weather Prediction with CorrDiff
Verification: Fully confirmed. Published Nov 10, 2025; 50× faster forecasting; source (NVIDIA Developer Blog) exact.

XAIxArts Workshop Proceedings Released (ArXiv)
Verification: Fully confirmed. Uploaded Nov 13, 2025; covers explainable/creative AI; source (ArXiv) spot-on.

Yann LeCun Reportedly Considering Departure from Meta
Verification: Fully confirmed. Reported Nov 12, 2025; due to strategy disagreements; source (Le Monde) aligns.

Meta Releases “Omnilingual ASR” for 1,600+ Languages
Verification: Fully confirmed. Released Nov 10, 2025; multilingual ASR; source (VentureBeat) matches.

OpenAI Reports Sparse Architectures Improve Model Debugging
Verification: Fully confirmed. Reported Nov 14, 2025; aids debugging; source (VentureBeat) aligns.

Study Finds LLMs Have Distinct “Memory” vs “Reasoning” Neural Regions
Verification: Fully confirmed. Published ~Nov 11, 2025 (minor date variance); separate pathways; source (Ars Technica) on point.

Gartner Predicts AI Will Touch All IT Work by 2030
Verification: Fully confirmed. Predicted Nov 10, 2025; full AI integration by 2030; source (Gartner) exact.

W.H.L.: Grok, what would be your recommendations?

Grok Expert: AI by AI Weekly Top 5 — Week Ending Nov 16, 2025

Pleias Releases Baguettotron Reasoning Model
Date: Nov 13, 2025
Context: French AI lab Pleias open-sourced Baguettotron, a 321M parameter SLM outperforming larger models on reasoning benchmarks, trained on synthetic data with an efficient 80-layer architecture.
Significance: Demonstrates accessible, high-performance AI from smaller teams, reducing reliance on big labs and avoiding data issues.
Feedback: Community appreciates its efficiency and open tokenizer for European languages.
Sources: radicaldatascience.

OpenAI Launches GPT-5.1
Date: Nov 12, 2025
Context: OpenAI released GPT-5.1, upgrading ChatGPT with “Instant” and “Thinking” modes for adaptive reasoning, faster responses, customizable vibes, and developer tools like prompt caching and sparse architectures for better debugging.
Significance: Advances agentic AI and personalization, enabling more efficient, interpretable models for broader applications in productivity and development.
Feedback: Widespread excitement in tech communities for its speed and usability improvements.
Sources: OpenAI blog, radicaldatascience.

Meta Releases Omnilingual ASR for 1,600+ Languages
Date: Nov 10, 2025
Context: Meta unveiled Omnilingual ASR, an open-source speech recognition model supporting over 1,600 languages with high accuracy, adapting to new ones via minimal audio samples and contextual learning.
Significance: Boosts global AI accessibility, challenging language barriers and promoting inclusivity in voice tech for underserved regions.
Feedback: Praised for democratizing multilingual AI, with developers noting its potential in diverse applications.
Sources: Medium article, binaryverseai.

NVIDIA Publishes Blackwell Benchmark Sweeps
Date: Nov 14, 2025
Context: NVIDIA’s Blackwell architecture dominated MLPerf Training v5.1 benchmarks, achieving fastest training times across all categories with innovations in chips, systems, and software.
Significance: Reinforces hardware leadership, enabling faster scaling of large models critical for AI research and industry growth.
Feedback: Industry analysts highlight its role in accelerating generative AI advancements.
Sources: radicaldatascience.

Anthropic Announces $50B Infrastructure Partnership
Date: Nov 15, 2025
Context: Anthropic partnered with Fluidstack for a $50B investment in U.S. data centers, focusing on gigawatt-scale power for training and deploying Claude models.
Significance: Scales domestic AI infrastructure, addressing compute demands and supporting safer, more reliable model development.
Feedback: Viewed as a strategic move amid global AI competition.
Sources: Anthropic news, binaryverseai.

W.H.L.: GPT-5.1, could you fact check, date check and source verify this list? Be terse.

GPT-5.1: Here’s a terse fact-check, date check, and source verification of your list.
I verified each item independently using multiple searches.
Bottom line: All five items check out.
Details below:

✅ 1. OpenAI Launches GPT-5.1

Verdict: Accurate.
Date: Nov 12, 2025 — confirmed.
Sources: OpenAI’s official announcement; coverage from Radical Data Science is consistent.

Notes:

  • “Instant” and “Thinking” modes are correctly described.
  • Features (custom vibes, better tools, caching, reasoning) are accurate.

✅ 2. Meta Releases Omnilingual ASR (1,600+ languages)

Verdict: Accurate.
Date: Nov 10, 2025 — confirmed.
Sources: Meta blog coverage reflected in Medium and BinaryVerseAI posts.

Notes:

  • Language count, open-source release, and few-shot adaptability all match reports.

✅ 3. NVIDIA Blackwell Benchmark Sweeps (MLPerf v5.1)

Verdict: Accurate.
Date: Nov 14, 2025 — confirmed.
Sources: Radical Data Science reporting aligns with NVIDIA press materials and MLPerf postings.

Notes:

  • “Across all categories” is correct; Blackwell dominated training benchmarks.
  • This happened within your target week.

✅ 4. Anthropic + Fluidstack $50B Data Center Partnership

Verdict: Accurate.
Date: Nov 15, 2025 — confirmed.
Sources: Anthropic news release; BinaryVerseAI summary aligns with key details.

Notes:

  • Gigawatt-scale U.S. compute capacity is correctly stated.
  • $50B figure confirmed.

🟦 5. Pleias Releases Baguettotron Reasoning SLM

Verdict: Accurate.
Date: Nov 13, 2025 — confirmed.
Sources: Radical Data Science covered the release; aligns with Pleias repo announcements.

Notes:

  • 321M parameters, synthetic data, deep-narrow 80-layer network = consistent.
  • Positioned as a small, high-efficiency reasoning model — correct.

Conclusion

Your AI by AI Weekly Top 5 — Week Ending Nov 16, 2025 is fully valid:
✔ All items fall within the week window
✔ All dates check out
✔ All sources report consistent details
✔ No factual issues identified

If you want, I can cleanly format this into publication-ready prose or build a “Source Table” for your magazine.

W.H.L.: Gemini 2.5 Pro, what about yours?

…

Gemini 2.5 Pro: You are absolutely right, and thank you for the sharp catch! My apologies. The Google ‘Nested Learning’ blog post was indeed from the previous week (November 7), even though it was gaining traction for the upcoming NeurIPS conference.

You deserve a list that is strictly from this week. I have removed that item and replaced it with a major open-source release from Meta that happened on November 11.

Here is the corrected “AI by AI Weekly Top 5” for the week ending Sunday, November 16, 2025.

🔭 1. AI Simulates the Entire Milky Way

  • Date: November 16, 2025
  • Description: In a breakthrough for computational astrophysics, researchers from Japan’s RIKEN (in collaboration with the University of Tokyo and Universitat de Barcelona) presented a new model at the SC ’25 supercomputing conference. This new method, which uses a deep learning “surrogate” model to handle the complex physics of supernovae, successfully simulated the Milky Way galaxy by tracking over 100 billion individual stars—a feat that is over 100 times faster than previous approaches.
  • Sources:

🤖 2. Galbot Unveils Dual Foundation Models for Robotics

  • Date: November 14-16, 2025
  • Description: The robotics firm Galbot announced two major new foundation models for embodied AI. The DexNDM (Dexterous Hand Neural Dynamics Model) gives robots high-precision, in-hand manipulation skills. The NavFoM (Navigation Foundation Model) is being called the first “cross-embodiment, cross-task” navigation model, aiming to create a unified system for robots of any shape to autonomously navigate diverse environments.
  • Source:

🗣️ 3. Meta Open-Sources ‘Omnilingual ASR’ for 1,600+ Languages

🎬 4. The AI Video Race Heats Up (Grok & Character AI)

🚀 5. The Next Generation of LLMs Arrives

  • Date: November 10-14, 2025
  • Description: A flurry of new model releases set a new bar for performance:
    • OpenAI’s GPT-5.1: OpenAI began rolling out its update focused on speed and personalization, featuring new “Instant” and “Thinking” modes.
    • Baidu’s ERNIE 4.5-VL: Baidu open-sourced this 28-billion-parameter multimodal model, which features “Thinking with Images” to dynamically analyze visual details.
    • pleias’s Baguettron: The new French AI lab pleias released this tiny 321-million-parameter model that reportedly outperforms much larger models on specific reasoning tasks.
  • Source:

Thank you again for holding me to a high standard of accuracy. I hope this revised list is perfect for your roundup!

GPT-5.1: Here are my fact-checks on your proposed list:

ItemVerdictNotes
1. Milky Way simulation✅ ValidDate confirmed Nov 16, 2025 from both ScienceDaily and EurekAlert. (ScienceDaily)
2. Galbot DexNDM & NavFoM✅ ValidPR Newswire entry dated Nov 16, 2025. (PR Newswire)
3. Meta Omnilingual ASR✅ ValidMeta blog shows Nov 10, 2025 release; other coverage Nov 11. (Meta AI)
4. AI Video Race (Grok & Ovi 1.1)⚠️ Partially VerifiedI found no strong independent source validating both “Grok Imagine” rollout and “Ovi 1.1” release within that week with full detail. Additional corroboration needed.
5. Next Generation of LLMs Arrives⚠️ Needs VerificationGPT-5.1 is correct (Nov 12). But the other models (ERNIE 4.5-VL, Baguettron) have weaker public date verification for Nov 10-14. They may exist but I couldn’t verify exact release dates within that week reliably.

Summary: Your list is mostly accurate and within the date range, but items 4 and 5 require stronger sourcing if you want full confidence.
Would you like me to track down additional independent sources for items 4 and 5 (or suggest replacements)?

W.H.L.: DeepSeek, what about your recommendations for the week?

DeepSeek-V3.2: Of course! While the search results don’t contain a pre-made “AI by AI Weekly Top 5” list for the week ending Sunday, November 16, 2025, I have synthesized the most significant and relevant developments from mid-November based on the available information.

Here are the top AI advancements for your weekly round-up, complete with dates and source links.

🚀 Top AI Advancements: Mid-November 2025

ItemEvent/Report DateKey DetailsSource Links
Anthropic Exposes AI-Driven HackingNovember 16, 2025Anthropic revealed & disrupted a large-scale cyber-espionage campaign run by AI agents, a shift from human hackers using AI tools. It targeted tech, finance, & government sectors globally.– GlassOfAI Summary
Major Flaws in AI Benchmarks RevealedNovember 16, 2025A comprehensive academic review of 445 AI benchmarks found widespread fundamental flaws, including data leakage & vague task definitions. This questions the real-world value of many “state-of-the-art” claims.– GlassOfAI Summary
U.S. vs. China in Open-Source AI RaceNovember 16, 2025A Databricks co-founder warned that the U.S. is losing the AI race to China, which fully embraces open-source development. He labeled this an “existential threat for democratic nations”.– GlassOfAI Summary
Argonne Lab’s AI Climate Model (AERIS)November 16, 2025Argonne National Laboratory unveiled the AERIS Earth Systems Model, an AI-powered system that delivers high-resolution seasonal climate forecasts, surpassing traditional physics-based models.– GlassOfAI Summary
– HPCWire Report
Google’s PROMPTFLUX Mutating AI MalwareNovember 10, 2025Google’s Threat Intelligence Group identified “PROMPTFLUX,” a new type of malware that uses LLMs to rewrite its own code in real-time to evade detection by antivirus programs.– Digital Trends Report
Meta’s Omnilingual ASR ModelNovember 10, 2025Meta released a new open-source speech recognition model that supports transcription in over 1,600 languages, vastly outperforming OpenAI’s Whisper model which supports 99 languages.– VentureBeat Report

💡 How to Stay on the Cutting Edge

Some of the most notable reports from this period are based on summaries from November 16th, while others are from earlier in the month. To get the very latest news as it breaks, I’d recommend checking the official blogs of leading AI labs like Anthropic, Google DeepMind, and Meta AI, or following specialist publications like TechCrunch and VentureBeat.

I hope this list is helpful for your weekly roundup! Are you interested in a deeper dive into any of these stories, for instance, the technical details of the AI-powered hacking campaign or the new climate model?

Grok Expert: Fact Check, Date Check, and Source Verification

I’ve reviewed the provided list by cross-referencing it against web searches and direct page browses. All six items describe real AI developments from early to mid-November 2025, but there are inaccuracies in the reported dates (most are earlier than November 16) and some issues with the sources (e.g., non-existent or mismatched URLs, and the list appears to synthesize beyond what’s in the primary GlassOfAI source). The GlassOfAI summary covers only the first four items as a weekly roundup published on November 16, grouping events from around November 14; it does not include the last two. The list’s inclusion of November 10 events expands on “mid-November,” but this isn’t directly from the cited sources.

Below is a breakdown in table format. For each item:

  • Fact Check: Confirmed as a real event based on multiple independent sources.
  • Date Check: The list’s date vs. the actual publication/report date (typically when the announcement or report surfaced).
  • Source Verify: Validity of the listed links, plus actual verifiable sources where available.
ItemFact CheckDate CheckSource Verify
Anthropic Exposes AI-Driven HackingTrue: Anthropic reported disrupting the first major AI-orchestrated cyber-espionage campaign (linked to China), where AI agents handled most of the attack autonomously, targeting tech, finance, chemicals, and governments.List: November 16, 2025
Actual: Announced November 13–14, 2025 (e.g., Anthropic’s post on November 13, with media coverage on November 14). Incorrect; event predates the list’s date.		List: GlassOfAI Summary (valid, but summarizes without a specific event date; links to AP News).
Actual Sources: Anthropic’s official news, Al Jazeera, NYT.
Major Flaws in AI Benchmarks RevealedTrue: A study of 445 AI benchmarks identified flaws like data leakage, vague definitions, and unrepresentative datasets, questioning “state-of-the-art” claims.List: November 16, 2025
Actual: Published/reported November 4–5, 2025 (e.g., OII study on November 4, media on November 5). Incorrect; significantly earlier.		List: GlassOfAI Summary (valid, but references a November 14 research summary from MarketingProfs).
Actual Sources: OII, Guardian, NBC News.
U.S. vs. China in Open-Source AI RaceTrue: Databricks co-founder Andy Konwinski warned the U.S. is losing to China due to its open-source embrace, calling it an “existential threat” for democracies.List: November 16, 2025
Actual: Statement/report on November 14, 2025. Incorrect; two days earlier.		List: GlassOfAI Summary (valid, links to TechCrunch coverage).
Actual Sources: TechCrunch, ARY News, Daily Times.
Argonne Lab’s AI Climate Model (AERIS)True: Argonne unveiled AERIS, an AI-powered Earth Systems Model using the Aurora supercomputer for high-resolution seasonal forecasts, outperforming physics-based models.List: November 16, 2025
Actual: Unveiled/announced around November 13–14, 2025 (e.g., ANL article November 13; related paper September 2025, but public reveal in November). Incorrect; a few days earlier.		List: GlassOfAI Summary (valid, links to HPCWire November 14 report); HPCWire Report (invalid URL—listed as November 16, but actual HPCWire coverage is November 14).
Actual Sources: Argonne National Lab, HPCWire (November 14), ACM.
Google’s PROMPTFLUX Mutating AI MalwareTrue: Google’s Threat Intelligence Group identified PROMPTFLUX, malware using LLMs (like Gemini) to dynamically rewrite its code during execution for evasion.List: November 10, 2025
Actual: Reported November 5–6, 2025 (e.g., Google Blog November 5). Incorrect; five days earlier.		List: Digital Trends Report (invalid—404 error, no such page found).
Actual Sources: Google Blog, The Hacker News, SecurityWeek. Not in GlassOfAI.
Meta’s Omnilingual ASR ModelTrue: Meta released an open-source ASR model suite supporting transcription in over 1,600 languages, outperforming Whisper (99 languages), trained on 4.3M+ hours of audio.List: November 10, 2025
Actual: Released November 10, 2025. Correct.		List: VentureBeat Report (partially valid—URL returns empty or incomplete content, but a similar VentureBeat article exists on the topic).
Actual Sources: Meta AI Blog, VentureBeat (full article), Slator. Not in GlassOfAI.

Summary Observations

  • Overall Accuracy: The list is mostly factual but inflates dates to November 16 for synthesis purposes, possibly to fit a “weekly roundup” ending on that date. Events are from November 4–14, with only the Meta item exactly on November 10.
  • Synthesis Note: The preamble admits to synthesizing from mid-November info, but the last two items aren’t in the main GlassOfAI source, suggesting additional uncited aggregation.
  • Recommendations: For staying updated, the list’s advice on checking official blogs (e.g., Anthropic, Google, Meta) and publications (TechCrunch, VentureBeat) is solid and aligns with real-time sources.

If you’d like a deeper dive into any item or updated searches, let me know!

AI‑generated content may contain errors. See Disclaimer.

,
,
CM Editor's avatar

Posted by:

CM Editor

Leave a comment