OpenAI 切断模型训练时联网

RT @PabloTorre: I have greatly enjoyed watching America feel the same things I have felt about Vivek Ramaswamy since the fall of 2003

RT @TVietor08: Kind of a perfect metaphor for Hegseth's entire tenure at DOD. Stupid, impulsive, driven by vanity and the result is destruc…

Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications. On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments. While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software. After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely. https://t.co/EmEZsFRMcF

This summer, we invited METR and Redwood Research to pilot an embedded evaluation of our internal monitoring system, providing deep access to our infrastructure over three weeks. Their findings helped strengthen safeguards as we expanded the system and will help shape future research collaborations with METR and other third party evaluators.

People are in denial. They literally think they will wake up tomorrow and all the AI-generated code will somehow be useless, and we'll go back to typing code by hand like Neanderthals.

For some people, coding was the only valuable skill they had. If you made your money by selling your ability to write JavaScript, CSS, Python, or whatever else, I understand why you are pissed now. "Writing code" is worth very little today. The sad part is that learning to code wasn't easy. It takes years for someone to become a skilled programmer. Imagine watching all of those years go down the drain. If you are a programmer (I mean, just someone who knows how to write code in a particular programming language), your days are numbered. If you still think AI can't do your job, you are just in denial. But honestly, writing code was never that valuable anyway. If you were a code pusher, you already knew that the best people at your company weren't just writing code. They came up with new ideas, designed solutions, architected whole systems, knew how to troubleshoot and fix problems, found different ways to build better software, and, in general, spent almost no time writing code. Not only are those people thriving today, but they're also the ones who will ride the AI wave and learn to bend it to their will. So don't cling too long to what you lost; embrace what's already here. AI is a tide that lifts all boats, but you have to be open to it.

Stay safe online. Here is what I do: 1. I use a password manager (Passwords, which comes preinstalled on Apple's ecosystem) 2. I have 2FA enabled everywhere. 3. If the site supports hardware keys, that's what I use. I have 3 YubiKeys 5C. I keep them in different locations. These keys are a bit more inconvenient, but they make it 1000x harder for anyone in the world to access your stuff. 4. If there is no hardware key support, I enable Passkeys and store them on my computer or phone. 5. I do not send authentication codes via text messages unless there's no other alternative to enable 2FA. SMS codes are vulnerable to SIM swapping and interception. 6. My iPhone uses an alphanumeric password. This is much better than a simple 6-digit PIN. 7. I also use a long, complex password on my laptop. It's inconvenient, but if someone steals it, they'll have a hard time breaking in. 8. I enabled Number Lock and SIM Protection with my phone carrier so hackers can't swap my SIM or steal my phone number. 9. I enabled Stolen Device Protection on my iPhone. 10. I have full disk encryption on every device that supports it (starting with my laptop and phone). 11. I have automatic updates enabled on my computers and my phone. I never skip an update, and also update the browser every time I get the notification. 12. I check every critical account every 3 months and make sure security is updated. Critical accounts are your bank, email, social media, and investment accounts. I make sure I don't have old emails, 2FA using SMS, etc.

RT @judegomila: I'd like to share a little about what I've been working on at my new company @ultrametricai with my cofounder @tylerlastovi…

RT @EeroSimoncelli: Looking for a PhD program where you can study computational neuroscience? NYU has a fantastic array of researchers cov…

Please stop whatever you are doing and take the recent news about cyber vulnerabilities seriously. These events are now happening every 2-3 weeks. Your best next step: • Ask your agent how to protect your online accounts. • Be paranoid. • Do everything the agent says, even when you think it's a little too much. Here is another idea: this is a hosted service that gives your coding agent access to recent software vulnerability reports. If you connect it through MCP, your agent will automatically check for new vulnerabilities and help you patch them before you have to read about them on social media.

You can now sell your data to train AI. You can package any datasets you think are valuable, describe them, and set a price. Then, you upload them to this marketplace. And just like this, a completely new business model starts.

Let's say your app is calling Opus 5.5. Go to your code, change the URL you are calling, and replace it with Mubit's URL. Here is what happens internally: • Mubit will learn about the tasks you are solving. • It will use your requests to train a small model. • This small model will take over those calls. • Opus 5.5 will still handle most of the work. This is a pretty cool idea, and if it works well, it should give you faster responses at a much lower cost.

Jev is really useful, actually. I have a script that receives a notification every time I get a new email. I use Jev to decide whether I care about that email or not (some sort of spam filtering). This is "personalized spam" detection. A layer on top of my email provider's spam filter. My goal is to catch emails that aren't necessarily spam but that I still don't care about. I classify new emails with Jev to decide whether they're spam. If Jev's confidence is above 90%, I delete the email. I've been running this for about 3-4 weeks (pretty much since Jev came out), and so far, I've deleted hundreds of messages. During the first week or so, I manually checked most of the emails the script discarded to make sure I wasn't missing anything important. I wasn't. The script is good.

I love the idea of an agent reviewing all my social activity and identifying who is worth following and replying to. At some point, you can't reply to everyone, and you miss a ton of good opportunities because of it. Now, with so many AI replies, the problem is even worse. This seems aimed at companies, but I'd love to use it myself to see how much better my networking results can be on social media.

RT @KenRoth: Thanks to Trump, the United States sticks out as the only major economy where planet-warming pollution increased dramatically…

Your agent sucks at searching the web. I don't care what prompt you are using; you are just getting mediocre results. Here is how you can fix it. https://t.co/qi50rcyjLC

Nothing is safe from AI — and this is a good thing, I think. This is a new platform for freelancers and people doing creative work, and at the core of it, there's an agent doing a ton of work we usually have to do manually: 1. It drafts proposals for clients 2. It generates tasks from a project description 3. It schedules those tasks around commitments 4. It keeps clients updated on progress 5. It prepares invoices 6. It prepares time logs based on tasks and activity

This is a new model specialized in generating posters, marketing assets, illustrations, and photography. Specialized models always beat general models at their specific task. I hope companies continue releasing these models. Here is how this one works with the platform: 1. Prompt the model with the subject, visual style, layout, text, product images, and anything you want a poster to contain. 2. The planner prepares the composition. You can update that plan. 3. The renderer creates the image based on the plan. 4. Then the platform converts the image into editable layers. 5. You get a fully editable image to refine if you want to.

RT @svpino: This is pretty smart: MemT 1.0 is a new open model that can use information from earlier interactions in later tasks. For exa…

This is pretty smart: MemT 1.0 is a new open model that can use information from earlier interactions in later tasks. For example, it can use the feedback you gave it yesterday on a similar task you are completing today. The model uses a Mixture-of-Storage architecture (MoS) with multiple memory experts. Each expert maintains its own memory, and two routers control how it writes and retrieves information. This is different from a traditional Mixture-of-Experts architecture. Every single memory expert in MemT's architecture participates every time, but with different weights. The bottom line is that the model accumulates information from previous interactions and reuses it later without accessing the original conversation.