replace_me
@replaceme.bsky.social
self.replace('manual', 'automated')
OpenAI used GPT to take "short user prompts" and generate "longer detailed captions" (synthetic essentially) to pair with videos as part of the training set, which enables Sora to generate high quality videos that accurately follow user prompts. openai.com/research/vid...
Video generation models as world simulators
We explore large-scale training of generative models on video data. Specifically, we train text-conditional diffusion models jointly on videos and images of variable durations, resolutions and aspect ...
openai.com
February 16, 2024 at 5:27 PM
OpenAI used GPT to take "short user prompts" and generate "longer detailed captions" (synthetic essentially) to pair with videos as part of the training set, which enables Sora to generate high quality videos that accurately follow user prompts. openai.com/research/vid...
The Needle In A Haystack (NIAH) for Gemini 1.5 Pro "found the embedded text 99% of the time, in blocks of data as long as 1 million tokens"
February 15, 2024 at 11:06 PM
The Needle In A Haystack (NIAH) for Gemini 1.5 Pro "found the embedded text 99% of the time, in blocks of data as long as 1 million tokens"
Google just announced Gemini 1.5 Pro with a stated context length of 1M tokens and a new MoE architecture (paper in blog). #llms blog.google/technology/a...
Our next-generation model: Gemini 1.5
Gemini 1.5 delivers dramatically enhanced performance, with a breakthrough in long-context understanding across modalities.
blog.google
February 15, 2024 at 3:47 PM
Google just announced Gemini 1.5 Pro with a stated context length of 1M tokens and a new MoE architecture (paper in blog). #llms blog.google/technology/a...
Interesting improvements achieved using hyperbolic inmage embeddings, very cool arxiv.org/pdf/1904.022...
February 9, 2024 at 5:32 PM
Interesting improvements achieved using hyperbolic inmage embeddings, very cool arxiv.org/pdf/1904.022...
What's the most useful LLM implementation you've either applied to your personal or work life? Mine has been LLM-generated commit messages (OpenCommit) based on staged changes to be committed. #llms #actuallyuseful
February 7, 2024 at 5:20 PM
What's the most useful LLM implementation you've either applied to your personal or work life? Mine has been LLM-generated commit messages (OpenCommit) based on staged changes to be committed. #llms #actuallyuseful
Reposted by replace_me
super excited to get this going
this proposal will allow you to securely login to any atproto application using your own self-owned, self-hosted identity
to be a bit grandiose, we hope it's the last account you ever need
this proposal will allow you to securely login to any atproto application using your own self-owned, self-hosted identity
to be a bit grandiose, we hope it's the last account you ever need
atproto devs, again!
are you an OAuth afficianto? how does that even work in a federated system, with mobile clients, single-page-apps, etc?
check out this big proposal, let us know your questions (here or as github issues)
are you an OAuth afficianto? how does that even work in a federated system, with mobile clients, single-page-apps, etc?
check out this big proposal, let us know your questions (here or as github issues)
github.com
February 6, 2024 at 8:41 PM
super excited to get this going
this proposal will allow you to securely login to any atproto application using your own self-owned, self-hosted identity
to be a bit grandiose, we hope it's the last account you ever need
this proposal will allow you to securely login to any atproto application using your own self-owned, self-hosted identity
to be a bit grandiose, we hope it's the last account you ever need
Who's writing about Transformers, LLMs, RAG and other data science topics on bluesky? #llms #rag #langchain
February 6, 2024 at 7:39 PM
Who's writing about Transformers, LLMs, RAG and other data science topics on bluesky? #llms #rag #langchain