Big if true!
Cool way to use Claude Code: deciphering Linear A, a 3500 year old written language from Crete
aiclambake.com/clamtakes/lin…
Hope this holds up in peer review! 🤞
1995: waiting for a JPEG to progressively load
2025: waiting for a 4o studio ghibli style image to progressively load
We will have achieved AGI when it can install CUDA and all the correct versions of whatever ML frameworks and tools you’re using on the first try without using a docker image. This is true alignment
More use of @liger_kernel @hsu_byron
Pytorch zoom backend: An experimental Triton first integration into PyTorch eager mode where the kernels are written in Triton (instead of CUDA or HIP). Uses Liger kernels now but can use any Triton kernel. Runs llamas. hack away at it.
Thoughts ?
github.com/nod-ai/pytorch/bl…
Great opportunity if you’re interested in performance optimization
We are hosting our 1st IRL event on 9/25 at the LinkedIn campus - "Scaling AI Infra - GPUs, Kernels, LLMs and More".
We will discuss liger-kernel and invite speakers to talk about DeepSpeed (@Guanhua_Wang_ ), SGLang (@ying11231), and the TensorCore (Pradeep Ramani, dePaul Miller) team. Please RSVP at the link:
scalingaiinfragpuskernelsllm…
Claude 3.5 Sonnet seems to have better build-in fact retrieval. My pet eval is to ask about a Glen Campbell album, after "priming" by asking "who was Glen Campbell".
Before (3.0) the output was almost entirely hallucinated:
I wonder if this is how advertising/promotion will be done. Clamp the features for product X, now LLM can’t stop talking about it.
This week, we showed how altering internal "features" in our AI, Claude, could change its behavior.
We found a feature that can make Claude focus intensely on the Golden Gate Bridge.
Now, for a limited time, you can chat with Golden Gate Claude: claude.ai
Even though California only had a partial eclipse, the effects on solar generation were clearly visible:
