About This Week’s Covers
This week’s cover is personal. For 67 weeks, I’ve organized AI headlines, averaging over 300 per week. I was behind for a long time. At one point, I was six weeks behind. I didn’t think I could catch up. I’ve wanted to quit. I’ve been embarrassed and fatigued. But I didn’t quit. This week, I hit 20,000 organized links, and I’m officially caught up. To capture the daunting task, energy, and triumph, I made a Flux LoRA of myself on a mountaintop, holding a sign that says: “Never Quit—20,000 Links.”
The rest of the covers were created using Claude 3.5 and the Ideogram API, with the related theme ‘summiting a mountain + category name.’
This Week’s Executive Summaries
AI Leaders Signal Unified View on Imminent AGI Breakthrough
Sam Altman, CEO of OpenAI, made waves in a new blog post claiming the company knows how to build artificial general intelligence (AGI) and predicts AI “workers” will transform companies as soon as 2025. Going further, Altman revealed OpenAI’s ambitions extend beyond AGI to developing superintelligent AI systems to dramatically accelerate scientific discovery and innovation. The post reflects a growing consensus among top AI lab leaders and researchers. Public statements and private discussions with experts suggest AGI development is advancing faster than previously expected.
https://blog.samaltman.com/
“Sam Altman’s just published new blog. “We are now confident we know how to build AGI as we have traditionally understood it. We believe that, in 2025, we may see the first AI agents “join the workforce” and materially change the output of companies.”
“This bit of Sam Altman’s newest post is similar in tone to a post by the CEO of Anthropic & what many (not all) researchers from every lab have been saying publicly and privately. You do not have to believe them, but I think they believe what they are saying, for what it worth.”
“i always wanted to write a six-word story. here it is: ___ near the singularity; unclear which side.”
NVIDIA Unveils Free Simulation Platform to Help Robots Better Understand the World
NVIDIA has released Cosmos, a new free toolkit that helps robots and self-driving cars learn about the physical world through virtual training. The platform uses AI to create detailed digital simulations of the real world, allowing machines to practice tasks safely before trying them in reality – similar to how pilots train in flight simulators. By making this technology freely available to developers, NVIDIA aims to solve one of robotics’ biggest challenges: gathering enough real-world training data. The platform comes pre-loaded with 20 million hours of video data and can generate synthetic training scenarios, potentially accelerating the development of more capable robots and autonomous vehicles.
DrJimFan | lifearchitect | reddit | mervenoyann | nvidia | huggingface
Google DeepMind Forms Elite Team to Build World-Simulating AI (sound familiar?)
Google DeepMind is launching a new initiative to create AI systems that can simulate physical reality, led by Tim Brooks, a former OpenAI Sora developer. The team aims to build on Google’s existing AI models (Gemini, Veo, and Genie) to create sophisticated simulations that could power everything from visual reasoning to robotic embodiment to interactive entertainment.
Techcrunch | _tim_brooks
As AI Gets Better, AI Research Firms New Tougher Tests to Measure Performance
After a two-year development period, the upcoming ARC-AGI-2 benchmark will launch in February 2025. The system measures how well AI systems perform on complex tasks.
“We’re going to be releasing ARC-AGI-2 towards late February. It’s been a long time coming — I first announced it in early 2022. Beyond that, we’re starting work on a next-generation AGI benchmark that completely departs from the 2019 ARC-AGI format. We’re very excited about it!” / X
“This tweet is exactly what you would expect to see in a world where AI capabilities are growing: last years very hard math tests built to challenge AI are already getting solved, we need much harder tests. Feels like the background news story in the first scene of a scifi drama.” / X
Multimodality is a big term to follow – agents, robots, multimodality!
For two years I’ve been convinced that segmentation and depth are a critical key to embodied robots. I’ve collected over 200 links about the topic. These are essentially object recognition and tracking tools. It’s wild to see the new DARPA challenge is all about computational imaging and ranging. Feels good to be on trend! In addition to the DARPA announcement, there were three other strong tools released this week which tie back to object tracking. I encourage you to check them out and try to understand how they will help robots in the future. What’s wild is they look like TikTok filters more than robot tools… but trust me.
“New DARPA challenge: Computational Imaging Detection and Ranging (CIDAR)”
To Interact With the Real World, AI Will Gain Physical Intelligence | WIRED
https://www.wired.com/story/ai-physical-intelligence-machine-learning
merve on X: “ViTPose — best open-source pose estimation model just landed to @huggingface transformers 🕺🏻💃🏻 See how to use on the next one ⤵️ https://t.co/lQYvT064Wu” / X – https://x.com/mervenoyann/status/1877360767478952012
“Released by @RealityLabs Research at #ECCV2024, Nymeria is a large-scale multimodal egocentric dataset for full-body motion understanding with potential applications in VR/MR headsets, smart glasses and more. More on this work + access to the dataset ➡️
I plan to keep a close record of these sorts of links (segmentation and depthing) and post a long article series on them in the coming months.
NVIDIA Revolutionizes Gaming Graphics with AI-Powered Rendering
NVIDIA CEO Jensen Huang presented a groundbreaking shift in computer graphics at CES 2025, announcing that their new graphics cards will use AI to generate over 90% of game pixels in real-time. The system works by using traditional ray-tracing to create a basic framework of about 10% of the image, while AI neural networks fill in the remaining details instantly. The technology also generates three additional AI-created frames for every traditionally rendered frame at 4K resolution. This marks a fundamental change in how video game graphics are created, leading to more detailed and efficient gaming experiences while requiring less raw computing power.
DrJimFan | rohanpaul_ai | rohanpaul_ai
It’s Time To Prepare for a LOT of Cultural Changes
“Working paper argues AI creates an “inflection point” for each job type. Before that, AI boosts freelancer earnings (web devs saw a +65% increase). After, AI replaces freelancers (translators saw -30% drop). They suggest that once AI starts replacing a job, it doesn’t go back.”
“Reminder for the new semester When researchers secretly added AI-created papers to the exam pool: “We found that 94% of our AI submissions were undetected. The grades awarded to our AI submissions were on average half a grade boundary higher than that achieved by real students.”
AI is weaving itself into the fabric of the internet with generative search | MIT Technology Review
https://www.technologyreview.com/2025/01/06/1108679/ai-generative-search-internet-breakthroughs
Elon Musk Launches Grok 2 AI app for iPhones, Announces Grok 3 Release Within a Month
(I guess he didn’t ban iPhones after all)
“IT’S HERE: xAI’s brand-new standalone Grok iOS app. Harness powerful AI, generate stunning images, and login with X to personalize your experience with real time news, sports and local data. Download now in the US:
“BREAKING: Elon Musk has just confirmed that Grok 3 will be released in 3-4 weeks.
“did i not tell you grok 3 is coming. by far the best model i’ve ever used and its not even close.” / X
AI Visuals and Charts: Week Ending 01/10/2025
“100% AI-generated dashcam footage. The line between reality and imagination continues to blur.
“We’re basically building the matrix inside the matrix, aren’t we?”
“Yup, we’re definitely building the matrix inside the matrix.”
“Allegedly shot in Shenzhen. Is this real? Can someone verify? I’ve seen this company posting very natural humanoid walking gaits a couple months ago. These days, it’s hard to tell CGI vs Sora vs real …
“Traditional VFX pipelines is dead. Adobe’s TransPixar unlocks real-time transparent effects with minimal training data It could redefine visual effects (VFX) production across film, gaming, and AR industries. → Developed by Adobe Research and HKUST, TransPixar integrates
“Increasingly large food. All made by me with veo 2, the serious point is to note how impressive the “physics” of these models have become (also the consistency, like the spilled burrito across two different shots), there are some issues, obviously, but big advances.
Top 39Links of The Week – Organized by Category
Agents and Copilots: AI News Week Ending 01/10/2025
“Browserless is a web service that allows remote clients to connect and execute headless browser tasks using Docker, supporting libraries like Puppeteer and Playwright, and offering REST APIs for functions like PDF generation and screenshot capture”
AGI (Artificial General Intelligence): AI News Week Ending 01/10/2025
o3: The grand finale of AI in 2024 – by Nathan Lambert
“Jared Friedman of Y Combinator says that while 2024 was the year you could have a natural-sounding phone call with AI, 2025 will be the year you can have a realistic video call with an AI
“@doomslide @chaitinsgoose I do not acknowledge this “disagree” because I agree: there is. What we might really be disagreeing about is who’s being pedantic. I think Sonnet 3.5 is meaningfully a proto-AGI, ie a big boy AGI can be derived through some relatively trivial (if costly) improvements on Sonnet” / X
Anthropic: AI News Week Ending 01/10/2025
Exclusive | AI Startup Anthropic Raising Funds Valuing It at $60 Billion – WSJ
“Anthropic reportedly in talks to raise $2B at $60B valuation, led by Lightspeed. This would bring Anthropic’s total raised to $15.7 billion. It would also make Anthropic the fifth-most valuable U.S. startup after SpaceX, OpenAI, Stripe, and Databricks. On a related note,
“@doomslide @chaitinsgoose I do not acknowledge this “disagree” because I agree: there is. What we might really be disagreeing about is who’s being pedantic. I think Sonnet 3.5 is meaningfully a proto-AGI, ie a big boy AGI can be derived through some relatively trivial (if costly) improvements on Sonnet” / X
Apple: AI News Week Ending 01/10/2025
Apple says it will update AI feature after inaccurate news alerts | Apple | The Guardian
Apple urged to withdraw ‘out of control’ AI news alerts
Apple opts everyone into having their Photos analyzed by AI • The Register
Apple in talks with Tencent, ByteDance to roll out AI features in China, sources say | Reuters
Augmented and Virtual Reality (AR/VR): AI News Week Ending 01/10/2025
“HOLY SHIT – generate 3D mesh from a single image in LESS THAN A SECOND 🤯
World Models
“World Foundation Models (WFMs) are essentially pre-trained on open-world video data, which you can then fine-tune it on your specific application with less labels (be it autonomous driving or robotic arms) This release matters so much for embodied applications because labelling
Chips, Hardware, and Infrastructure: AI News Week Ending 01/10/2025
NVIDIA
“We’re so unfathomably back – @NVIDIAAI releases Cosmos: World foundation models (commercially permissive) 🔥 Models trained on over 20 MILLION hours of video can be used to generate dynamic, high quality videos from text, image, or video inputs 🤯 Available directly on Hugging
Ethics/Legal/Security: AI News Week Ending 01/10/2025
Elon Musk says all human data for AI training ‘exhausted’ | Artificial intelligence (AI) | The Guardian
Apple opts everyone into having their Photos analyzed by AI • The Register
Biden Administration Ignites Firestorm With Rules Governing A.I.’s Global Spread – The New York Times
ByteDance appears to be skirting US restrictions to buy Nvidia chips: Report | TechCrunch
AI Chip Curbs Trigger Rare Public Fight: Tech Giants vs. China Hawks – WSJ
“This AI has NO rules. Hackers use it to write perfect phishing emails and generate deadly malware in seconds. It’s like ChatGPT—but evil. How do you stop something built for destruction?”
Google: AI News Week Ending 01/10/2025
Code Assist, Google’s enterprise-focused coding assistant, gets third-party tools | TechCrunch
Google’s CEO warns ChatGPT may become synonymous to AI the way Google is to Search
Google unveils an AI-powered TV that summarizes the news for you at CES 2025 | TechCrunch
OpenAI: AI News Week Ending 01/10/2025
Reflections – Sam Altman
Introducing ChatGPT Pro | OpenAI
“insane thing: we are currently losing money on openai pro subscriptions! people use it much more than we expected.” / X
“New study on AI & investing: When GPT-4o summarizes earnings calls to match investor expertise level (simpler for novices, technical for experts): Sophisticated investors get +9.6% improvements in 1-year returns, novices: +1.7% AI helps everyone, but expertise amplifies benefits!
“OpenAI may launch agents this month They were lagging because of concerns of prompt injection attacks. According to ‘TheInformation’ —— Prompt Injection Attacks involve manipulating the input (or “prompt”) to alter its behavior in unintended or malicious ways. So
“Build a multi-agent AI news assistant using Open AI swarm and Llama 3.2 running locally on your computer (100% free and open source):” / X
‘Virtual employees’ could join workforce as soon as this year, OpenAI boss says | Technology sector | The Guardian
Publishing: AI News Week Ending 01/10/2025
Apple says it will update AI feature after inaccurate news alerts | Apple | The Guardian
Apple urged to withdraw ‘out of control’ AI news alerts
“Browserless is a web service that allows remote clients to connect and execute headless browser tasks using Docker, supporting libraries like Puppeteer and Playwright, and offering REST APIs for functions like PDF generation and screenshot capture”
“Pushed the initial subsets of a BBC News curated @huggingface FineWeb dataset overnight, expecting ~400M entries total ⏳⏳⏳ 💾
Robotics and Embodiment: AI News Week Ending 01/10/2025
NVIDIA
Nvidia unveils robot training tech, new gaming chips and Toyota deal | Reuters
“Robotics has a data scarcity problem – you simply can’t scrape robot control data from webpages. Introducing GR00T-Mimic and GR00T-Gen: using both Graphics 1.0 & Graphics 2.0 to multiply your robot datasets by 1,000,000x. We trade compute for synthetic data, so we are not capped
“NVIDIA’s greatest strength is that literally the same software stack (CUDA, cuDNN, NCCL, TensorRT, Isaac) runs on both xAI’s 100K beast of a cluster and humanoid robot’s tiny silicon heart. We do lots of R&D in all verticals, because we as a company needs to understand what
“People ask me what’s next. The GOAT points the way. Physical AI. Embodied Agents. Robotics. That’s what’s next.
Video News: AI News Week Ending 01/10/2025
A new, uncensored AI video model may spark a new AI hobbyist movement – Ars Technica
Leave a Reply