Gemini Diffusion
Test Information Space
(2) Erik Brynjolfsson on X: "A few months ago, the best LLM scored 5% on the USA Math Olympiad test. Models have been rapidly improving. Today, Google Gemini 2.5 scored 49%, which is better than 75% of the people who took the test (roughly the top 250 students in the USA)." / X
(If not upcoming: DOGE test.)
Scalable Chain of Thoughts via Elastic Reasoning
Reward Reasoning Model
Think Only When You Need with Large Hybrid-Reasoning Models
View PDF
Understand all the I/O news with NotebookLM.
Google I/O 2025: From research to reality
When Thinking Fails: The Pitfalls of Reasoning for...
View PDF
Questioning Representational Optimism in Deep Learning: The...
View PDF
On the conversational persuasiveness of GPT-4 - Nature Human Behaviour
Google launches NotebookLM for Android
Generalization bias in large language model summarization of scientific research | Royal Society Open Science
Moihosso Portable Charger Power Bank 56800mAh Battery Pack-22.5W Fast Charging with USB C & LED Display, 3 Outputs High Capacity Portable Power Bank for iPhone 16/15/14/13 Series, Samsung, iPad etc : Cell Phones & Accessories
(USB batteries are economical during the day where off-peak charging is lower cost, e.g. 20% less, overnights and weekends.)
Forget Humanoids. At MIT, Worms and Turtles Are Inspiring a New Generation of Robots
Google is about to unleash Gemini Nano's power for third-party Android apps
(Big news. People trained for administrative hierarchies that faced networks on the ground will now have these edge apps. The former was not social networking. Now whoever gets shuffled out to the periphery and looks like a researcher can actually be oriented and have the tools. Franklin used to wonder how something would be of most use? For instance, nice to get points for trying, or vibe coding, the most games. Also orient to crypto economics. What flips the theory? Or philosophy? Natural yields to synthetic in more cases. What is the data crop? Whatever the iteration, they are all likely to beg the question. Industry and frugality. AI orgs have been arguing for speed and scale. Those become a framework in which to rank or position competitors. In contrast to the leaderboard. Enjoy while it lasts since there is the original flavor of nano-tech out there. Meanwhile, if regulators or the public seem too negative, at least this offers options. Franklin could also adapt the pro/con algebra from any perspective which was a strategic negotiating advantage; the Proxy of its day. One would expect templates, if not cover, designs for those native journals and travelers notes. Here comments stayed on Raindrop for AI, others on AmpleNote, but the clock is ticking.)
Introducing Codex | OpenAI
Recipes For Reasoning, Open and Compact Code Generator, Looser AI Regulations, and more...
(Cons reported elsewhere: LLMs are costlier than databases, harder to comprehend result ranking, and tricky for air-gapped datacenters. Relate to Primer report on Semantic Search. Nonetheless the current government admin is proposing to merge 3 dozen databases. Potentially accessible to DOGE.)
Exclusive: Legal startup Harvey AI in talks to raise funding at $5 billion valuation
The Middle East Has Entered the AI Group Chat
What to Look Out For When Acquiring AI Systems
Elon Musk Predicts Robots Will Unlock “Universal High Income” at Saudi Tech Event
(3) OpenAI on X: "By popular request, GPT-4.1 will be available directly in ChatGPT starting today. GPT-4.1 is a specialized model that excels at coding tasks & instruction following. Because it’s faster, it’s a great alternative to OpenAI o3 & o4-mini for everyday coding needs." / X
AlphaEvolve: A Gemini-powered coding agent for designing advanced algorithms
When it Comes to AI Policy, Congress Shouldn’t Cut States off at the Knees
New Google Cloud certification in generative AI | Google Cloud Blog
Philosophy Eats AI: What Leaders Should Know
(Actually tried this with Thema on Quora aimd creeping expansion before they threw up their hands and settled on natural search and summary. They are showing traffic tides on POE as well as multimodal builds out. Suspect the Proxy will strike back as scientific model and hint of what is yet to be known. AI has its own specializations and spinoffs which may outnumber philosophy's before long. Therefore may need a method to wrangle between the pair. In any case the dialogue would include the UK at the kickoff, Europe, Russia, China, Mideast, and so on, on their Marx, eventually to whatever colonizes Space or the emerging domains. Unless a Vulcan ambassador appears. The early agenda was to rediscover the lost continent of cognition. What eats information? Or to torture a trope, what is the transformer party drinking?)
Report: Spring 2025 AI Model Usage Trends - Poe
(3) Robert Scoble on X: "It's not every day I get to interview a former principal scientist who worked at Google, and is a Professor Emeritus at Stanford University, about the state of AI. But here we go. Introducing an hour with Yoav Shoham, @yshoham, AI pioneer and cofounder of @AI21Labs . This https://t.co/cawCtDwnSa" / X
Introducing Notion AI for Work
Evaluating AI Systems - Trends in AI: May '25