搜索“OpenAI”

anyone in the world. The next 12-18 months might be rough, as offensive capabilities arrive before patches. This was all predicted by the Situational Awareness paper, from former OpenAI employee Leopold Aschenbrenner. He said there is a scary time in between where attacker will have the upside, and it seems we are entering this period. Conclusion With all of this actually comes a great opportunity, which is that many of the core projects that run the Internet and computing systems are going thr

匹配字段正文

their pricing, their availability, their release cadence, and their tradeoffs. You're also missing what's happening outside of them. We've already seen how fast the ground shifts. OpenAI closed the gap faster than anyone expected; Google entered with competitive performance and dramatically different pricing; Benchmarks have tightened to the point where the differences between top models are increasingly hard to separate Teams are also paying closer attention to headlines around supply chain ris

匹配字段正文

了 AI 智能带来的好处。那么，更好的界面是什么样子？ 一种选择是为特定工作或任务构建专门的界面。在所有专门的 AI 界面中，真正完整的只有编程类界面。这完全在预料之中：AI 实验室的员工是程序员，模型在代码上接受了大量训练，而构建这些工具的人往往是为自己而建。 我之前写过 Claude Code，这是 Anthropic 的编程智能体，可以自主工作数小时。OpenAI 的 Codex 和 Google 的 Antigravity 也能做类似的事情。我用 Claude Code 做过各种事情，从赚（少量）钱到做游戏，完全没碰过任何代码。我也发现 Codex 极其有用，能力水平相近。这些工具非常出色，但它们确实是为程序员打造的。它们假定你懂 Python 和 Git。它们的界面看起来像 1980 年代的计算机实验室。对于并非开发者的 99% 知识工作者来说，这些强大的 AI 工具并没有为他们优化。 Pomelli、Stitch 和 NotebookLM 在所有 AI 实验室中，Google 似乎在为其他职业构建专门界面方面试验得最多。它们都还有些粗糙，但展示了当 AI 工具为其他类型的知

匹配字段正文

ds the ceiling to keep rising. These are very different pressures, and time favours the latter. The obvious counter-argument: if specialisation is where the value lives, why can’t OpenAI or Anthropic or Google just offer fine-tuning? They already have fine-tuning APIs. What stops them from capturing the specialisation layer too? The answer is that capturing the specialisation layer requires a fundamentally different organisational shape. Training a single frontier model is a concentrated effort:

匹配字段正文

em. I hate every other LLM framework Seriously. The way they implement LLM objects is painful. So I wrote my own. Super easy way to do calling. That's all they do - for Anthropic, OpenAI, and Google. Based on our telemetry, these account for 95% of use cases. Same interface. Full control over caching, serialization, provider quirks. No magic. No surprises. It's so much easier to do caching and implement messages yourself. Completely model-agnostic. You're not locked into one provider. You just d

匹配字段正文

t to cover, but let’s start chronologically in January 2025. Scaling still worked, but it didn’t really change how LLMs behaved or felt in practice (the only exception to that was OpenAI’s freshly released o1, which added reasoning traces). So, when DeepSeek released their R1 paper in January 2025, which showed that reasoning-like behavior can be developed with reinforcement learning, it was a really big deal. (Reasoning, in the context of LLMs, means that the model explains its answer, and this

匹配字段正文

升。 作为这一特定噩梦的预演，我给 Claude 一份公司备忘录，让它把它变成一份 PowerPoint。然后又从另一个角度做一份 PowerPoint。再做一份。 直到我得到了17份不同的 PowerPoint。那 PowerPoint 太多了。 如果我们不认真思考我们为什么要做工作，以及工作应该是什么样子，我们都会淹没在AI内容的浪潮中。替代方案是什么？OpenAI的论文提出，专家可以与AI合作解决问题，把任务交给AI作为第一遍处理，然后审查其工作。如果结果不够好，他们应该尝试几次给出修正或更好的指令。如果这仍然行不通，他们就应该自己动手完成工作。如果专家遵循这一工作流程，论文估计他们完成工作的速度会快百分之四十，成本会低百分之六十，更重要的是，还能保持对AI的控制。 智能体已经到来。它们能完成真实的工作，虽然这些工作仍然有限，但很有价值，而且在不断增加。然而，同一种能在几分钟内复现学术论文的技术，也能生成17份没人需要的PowerPoint演示文稿。这两种未来之间的差别不在于AI，而在于我们选择如何使用它。通过在决定什么值得做、而不只是什么能做时运用我们的判断力，我们可以确保这些

匹配字段正文

e-coherent cluster (setting apart Google, master of multi-datacenter-training). However, Colossus 1’s 300 MW looks modest next to the Gigawatt-scale clusters under construction by OpenAI, Meta and Anthropic. Their hyperscaler partners are happy to leverage their balance sheet and win the market by throwing dollars at it. Was xAI’s prowess a one-time wonder? Today we will publicize some data from our industry leading datacenter model over the last year that is accessible to clients. This is the o

匹配字段正文

g a 72-logical GPU with all to all switched scale-up size just like the VR200 NVL144) design similarly can lean on Meta internal inference workloads. Even emerging designs such as OpenAI’s chip with Broadcom stand to compete given they will be codesigned with Frontier Models in mind and with an internal workload as a backstop. Despite enjoying the benefit of internal demand, the MTIAv3 is excluded from this category due to its small 16 GPU world size. It will effectively now need to develop pref

匹配字段正文

to produce a large multimodal model, primarily focusing on text only due to compute limitations. They will release multimodal models this year, but they are still squarely behind OpenAI, Anthropic, and Google in many capabilties. If Blackwell GPU shipments come too early, that will accelerate China's pace. The second major implication of this is that China will now have more compute to serve and inference models to its population. Compute constraints affect user experience, with DeepSeek delibe

匹配字段正文

iplying revenue fivefold year-to-date to reach $5B annualized. To keep that trajectory, Anthropic is betting hard on Scaling Laws. While Dario’s startup draws fewer headlines than OpenAI, xAI and Meta Superintelligence, it isn’t shy about investment. AWS has well over a gigawatt of datacenter capacity in final stages of construction for its anchor customer. AWS is building datacenters faster than it ever has in its entire history. And there’s much more on the horizon. Trainium vs GPUs While Amaz

匹配字段正文

on weekly users. Gemini and other leading AIs add hundreds of millions more. In my posts, I often focus on the advances that AI is making (for example, in the past few weeks, both OpenAI and Google AIs chatbots got gold medals in the International Math Olympiad), but that obscures a broader shift that's been building: we're entering an era of Mass Intelligence, where powerful AI is becoming as accessible as a Google search. Until recently, free users of these systems (the overwhelming majority)

匹配字段正文

nto the future of HBM. We will cover the revolutionary change coming to HBM4 with custom base dies for HBM, what various different accelerators are doing with custom HBM including OpenAI, Nvidia, and AMD, the shoreline area problem, memory controller offload, repeater PHYs, LPDDR + HBM combos, and various beachfront expansion techniques. We will also discuss SRAM tags, compute under memory, supply chain implications, and Samsung. A Brief Overview of HBM As AI models grow in complexity, AI system

匹配字段正文

$200 million over 4 years. That is 100x that of their peers. Furthermore, there have been some billion dollar offers that were not accepted by researcher/engineering leadership at OpenAI. While these offers aren't all successful, Zuck is crushing the competitors by drastically increasing their cost per employee. Perhaps even more iconic, Zuck threw his entire Datacenter playbook into the trash and is now building multi-billion-dollar GPU clusters in “Tents”! As this report details, nothing is of

匹配字段正文

very important. The sooner the world can start a conversation about what these broad bounds are and how we define collective alignment, the better We (the whole industry, not just OpenAI) are building a brain for the world. It will be extremely personalized and easy for everyone to use; we will be limited by good ideas. For a long time, technical people in the startup industry have made fun of “the idea guys”; people who had an idea and were looking for a team to build it. It now looks to me lik

匹配字段正文

algorithm that only excels in toy settings. So why don’t we first figure out the environment we actually want to solve, then find the algorithm best suited for it? That’s exactly OpenAI’s initial plan. It built gym, a standard RL environment for various games, then the World of Bits and Universe projects, trying to turn the Internet or computer into a game. A good plan, isn’t it? Once we turn all digital worlds into an environment, solve it with smart RL algorithms, we have digital AGI. A good

匹配字段正文

{\url{https://metr.org/blog/2025-01-17-ai-models-dangerous-before-public-deployment/}} , year = {2025} , month = {01} , } Many frontier AI safety policies from scaling labs (e.g. OpenAI’s Preparedness Framework, Google DeepMind’s Frontier Safety Framework, etc.), as well as past work by third party evaluators including UK AISI , Apollo Research , and METR , focus on pre-deployment testing – ensuring that the AI model is safe and that the lab has sufficient security before the lab deploys the mo

匹配字段正文

any has won, and network effects have become self-perpetuating and insurmountable, then you don’t beat that by making the same thing but slightly better, and getting a judge to give you an entry point. You win by making the old thing irrelevant. Google didn’t build a better PC operating system or a Win32 office suite, Facebook didn’t do better search, and Apple didn’t do a better Blackberry. And OpenAI got 100m users in two months without going to a judge. Are better models better? The AI summer

匹配字段正文

ted numbers of people had both heard of it and used it. As with every observation about the acceleration of tech adoption, a lot of this is ‘standing on the shoulders of giants’ - OpenAI didn’t have to wait for people to buy devices or for telcos to build DSL or 3G. For consumers, ChatGPT is just a website or an app, and (to begin with) it could ride on all of the infrastructure we’ve built over the last 25 years. So a huge number of people went off to try it last year. The problem is that most

匹配字段正文