OpenAI 升级图像生成，Sam Altman 谈 AI 控制权

RT @PabloTorre: What is the point of a politician — let alone a literal vice president — who says he understood all of this and yet did abs…

RT @minakimes: Been a rough 24 hours huh

RT @thsottiaux: Day 25/ We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today. We heard yo…

RT @paulg: Now we'll finally learn what's in the Epstein files. Mike Johnson is no longer in a position to shut down the House in order to…

Images, upgraded—introducing ChatGPT Images 2.6. Now with GIFs and reference images. Plus, two new models are coming to the API with 50% faster image generation. https://t.co/kcuOhTipYR

People should always be in control of the future. OpenAI is against the notion that it would be acceptable for people to be relegated to a role of doing things that “AI systems think makes sense to reward in humans”. People should not cede control of the future to superintelligence, and that requirement should ultimately constrain AI progress.

Introducing GPT‑6 Cyber: a major advance in security research that pairs stronger capabilities with better model alignment. The model powered the Linux kernel research we announced yesterday, and security teams at companies like Mozilla, Palo Alto Networks, Calif, and Cisco—alongside our own at OpenAI—are already putting it to work. GPT-6 Cyber is available to select Daybreak Red customers and we are expanding access over time. https://t.co/Nzab3Z1Td6

Our security researchers built an autonomous cyber vulnerability discovery, exploitation, and patching pipeline. In a single run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million.

RT @SteveRattner: Democrats have won more votes than any party in midterm history: 66 million and counting. https://t.co/7s6vBjTUzG

RT @ylecun: @heyshrutimishra LLMs accumulate a lot of declarative knowledge. That's how they can answer a large number of questions. That'…

RT @richarddeitsch: That CBS number is incredible — and by incredible and I mean spectacularly and catastrophically awful. They should hav…

RT @SteveRattner: Elon Musk dropped nearly half a billion dollars on the midterms. 98% of the money he spent on specific candidates went…

RT @joshtpm: every couple days for the last few weeks I've watched this for like a hundred times straight.

RT @OfficialLoganK: Today we are rolling out Gemini Omni Flash 1.2, with: - longer video generation (15 seconds) - deeper cinematic contr…

RT @AlexiGlad: looped reasoning is completely overrated Energy-Based Reasoning (EBR) runs circles around looped reasoning paper dropping…

RT @Arjunjain: I was trained to think Yann LeCun was wrong. My PhD was in Vapnik’s school. Kernels. Bounds. Neural nets were alchemy. The…

RT @ylecun: @Arjunjain Plot twist: I was Vapnik's Department Head between 1996 and 2002, and made sure he had the space, resources, and col…

RT @reidhoffman: Narrative violation? https://t.co/N9TB8ycsgL

The era of System One Vision Model is here! Introducing VFlash: parsing the visual world in parallel at unmatched speeds for Physical AI. Most embodied prediction tasks like object localization, robot trajectory planning, and dense 3D motion are fixed-length float vectors. For example, a bounding box is always 6 tokens, a 10-waypoint robot arm trajectory is 54: we know these beforehand. Instead of decoding one token at a time, we can output the entire geometric shape in one go. VFlash enables up to 12x faster action decoding and a 5.1x overall robot inference speedup on edge GPUs like the Jetson Thor and DGX Spark. What’s surprising is that it ends up even more accurate than the VLM backbone, with up to 61% lower 3D trajectory error. The intuition is that VFlash avoids the cascading errors of autoregressive decoding, where every coordinate is written in ink and every later point loses the chance to backspace. Everything that moves will be autonomous, and everything autonomous will need to think on its own chip. The next inference revolution is physical reasoning at the speed of reflex. And as always, it’s open paper, open source! Link in thread:

RT @antirez: Remember one thing when you read JEPA vs LLMs in the context of the platonic representation hypothesis. Hidden state of an LLM…

RT @daveeyyy_: Yann LeCun: The Al industry is stuck because every lab is chasing LLMs. Real Al agents will need to predict the consequences…

RT @sean_n_wu: The Collaborative Intelligence Benchmark paper is out! We tested 18 AI models as teammates on 240 tasks. The best model sco…

RT @AndrewYNg: Join me and 2000 other AI engineers and builders at AI Dev in New York City, Nov 30-Dec 1! A LinkedIn study found that over…

RT @KenRoth: This is the current NY Times forecast. Incredible! https://t.co/Su7Kuce3Lb

RT @GaoyueZhou: Introducing ID-WM: a hierarchical world model that decouples "where to go" from "how to get there." Finally, we can plan ov…

You can now add a talking animated character to a website or app. Here is the API that will let you do that. You take an image and a short voice recording, and it turns them into an avatar you can talk to live.

RT @fouadmatin: Today’s GPT-6 Cyber release is a meaningful step for defenders: stronger vulnerability research and better alignment. We’ve…

Companies are killing for good engineers right now! These guys are offering a paid first-week vacation worth up to $10,000, planned with a dedicated travel concierge for new engineers who join their team. If you refer a candidate who joins, they give you a $30,000 bonus, and the person who joined also receives one.

Open models keep getting better! You can run this 8.8B-parameter model on your hardware, and it comes with impressive benchmarks, including 93% of frontier performance on coding and 92% on instruction following. @thewebAI cooked here. The link to the Hugging Face model page is below.

RT @che_shr_cat: 1/ A 109M text model trained with good views matched next-token prediction with 42x fewer samples on WIT-1024 (.109 at 21.…

RT @Brian_Bo_Li: @ylecun is certainly one of the few true thinkers and innovators in today’s overhyped AI landscape. He is brave to say s…

RT @svpino: Redis is cutting down the cost of running AI applications: 1. Semantic caching — Reuses generated answers from previous reques…

This is a shared memory service where you can connect multiple agents and AI tools to reuse knowledge. For example, you can save information using Codex and retrieve it through Claude Code. Two scenarios where this is very helpful: • To keep a portable knowledge base not tied to any agent. • To have the flexibility of using multiple agents without duplicating the knowledge.

Agents don't fail because models are dumb. They fail because they don't have the right context. Redis just released 6 new features for Iris, their agent context and memory platform: 1. Intent Router: Routes requests by meaning 2. Agent Playbooks: Reuses answers and procedures 3. Agent Memory: Remembers context across conversations 4. Context Gateway: Connects agents to data and tools 5. Lens: Traces agent decisions 6. Agent Mesh: Keeps agents communicating through interruptions Link to more info about this release: https://t.co/E9i4kFOLDZ

RT @KenRoth: In the United States, Brazil, Hungary, Serbia and Israel, "voters are turning out in droves to cut autocrats down to size. In…

RT @shaunbjohnson: At @224ventureshq, one thing we look for when we speak to founders is speed of execution. Execution is critical because…

This is a desktop agent that connects video generation models with creative tools such as Blender, DaVinci Resolve, and Premiere Pro. The idea is simple: Tell it what you want, and the agent will generate footage, assemble it, and edit it using pro tools.

RT @randall_balestr: If you are at INFORMS in San Francisco, come by room 211 where I will be talking about our latest JEPA research in 10m…

Building more software means managing more data, access rules, deployments, and monitoring. Every company I talk to is using agentic coding to do more. Codebases are growing, and so are problems. Here is the best way to manage complexity: ↓

Wordle 1,964 5/6 ⬛🟨⬛⬛⬛ ⬛⬛🟨⬛⬛ 🟨⬛🟨🟨⬛ 🟨🟨⬛⬛🟨 🟩🟩🟩🟩🟩

How to replace a large, expensive, and slow model with a small, cheaper, fast model and still get the same results. 7x faster and 50% fewer tokens. This is like fine-tuning, but without touching the model. Here, we teach Haiku, using Zep Agent Skills, to do what Opus 5.5 does to complete a task. • It takes 7 minutes with Opus 5.5. Only 59 seconds with Haiku. • It takes 1.4M tokens with Opus 5.5. Only 700K with Haiku.