The latest blog posts to take off on Hacker News. 673 posts.
Our commitment to developing cutting-edge open models in multiple modalities necessitates a compute solution capable of handling diverse tasks with efficiency. To this end, we conducted a performance analysis, training two of our models, including the highly anticipated Stable Diffusion 3.
The demand for computing power in the AI era is increasing at an unprecedented exponential rate. Luckily, for the past several decades, the miniaturization of CMOS transistor technology following Moore’s law has allowed much of this exponential growth to be accounted for by increasing computer efficiency.
LLMs seem surprisingly good at many things. So much so that not a week goes by without someone coming up with yet another use-case for this technology, often to solve tasks quickly that traditionally took a non-trivial amount of human work to complete. Today’s example was Brendan Dolan-Gavitt’s remarkable experiment…
We’re releasing an open source system, based on FSDP and QLoRA, that can train a 70b model on two 24GB GPUs.
Introducing a series on AI vs. ads
Bard IDOR, a GraphQL directive-overloading DoS that hung Google's backend for 109 seconds, and Workspace data exfiltration via markdown images.
Chronicles of training strong LLMs from scratch in the wild
Cloudflare is one of the first providers to safeguard LLM models and users in the era of AI
Following our announcement of the early preview of Stable Diffusion 3, today we are publishing the research paper which outlines the technical details of our upcoming model release, and invite you to sign up for the waitlist to participate in the early preview.
BASED is a linear attention architecture that decodes without a KV-cache, delivering a 24x throughput improvement over Transformers with FlashAttention-2.
This 1.5 release is unlike any of the previous ones. It brings many new features that can improve quality and the general audio experience. That is achieved through machine learning. Although Opus has included machine learning — and even deep learning — before (e.g. for speech/music detection ), this is the first time…
The Claude 3 model family—Haiku, Sonnet, and Opus—sets new industry benchmarks across a wide range of cognitive tasks.
AI, science, education, & ethics
Marketing, solicitations, advertising, political contact, direct mailers, email campaigns, cookies, ads — it's all opt out. It puts the burden on you to opt out of each communication each message, each outreach from a loyalty program. It can feel like weeding a garden, but that garden is your attention. The weeds keep…
Physics likes optimization! Subject to its boundary conditions, the time evolution of a physical system is a critical point for a quantity called an action. This point of view sets the stage for Noether's principle , a remarkable correspondence between continuous invariances of the action and conservation laws of the…
Nobody wants “temperamental and mysterious” products, so what will AI-integrated software development look like?
What are ML artifacts?
I think I’m probably going to lose quite a lot of money in the next year or two. It’s partly AI’s fault, but not mostly. Nonetheless I’m mostly going to write about AI, because it intersects the technosphere, where I’ve lived for decades.
Today, I’m going to write a bit about our GPU cluster machine we call cogito at Driveline Baseball. I’m going to timebox these blog posts so I get them out and also save my sanity, so t…
Corruption everywhere, even in YouTube's kids content
This post is based on my project in my Computer Vision class last semester
The Google Gemini fiasco shows that the biggest challenge for Google in AI is not business model but rather company culture; change is needed from the top down.
The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.
The State Space Model taking on Transformers
Modern AI rewards being very weird.
A complete GPT2 implementation as a single SQL query in PostgreSQL.
Meta's TestGen-LLM is a sneak peek to the future of developer productivity: specialized, orchestrated, and rigorously filtered.
An overview of the big ideas behind Mamba, a brand-new language model architecture.
