Some mockups I made a while back as part of Loop publication. #design #graphicdesign #sustainability

Show Post

Salesforce AI open-sources (code, model, dataset, and paper) BLIP3-o, the unified multimodal models that excels at both image understanding and generation in a single autoregressive architecture! It also includes: ▶️ Complete model implementations ▶️ Model weights

Show Post

But concerns exist about wealth inequality. Will the benefits of LLM-controlled devices be widely distributed, or will this tech primarily serve the wealthy, widening the gap? A critical societal debate. 6/6

Show Post

Long videos are a nightmare for language models—too many tokens, slow inference. ☠️ We explain STORM ⛈️, a new architecture that improves long video LLMs using Mamba layers and token compression. Reaches better accuracy than GPT-4o on benchmarks and up to 8× more efficiency. 📺 youtu.be/uMk3VN4S8TQ

Show Post

One of the biggest bottlenecks in inferring a large language model (LLM) is memory bandwidth, as the model's weights need to be constantly transferred between the GPU and memory. But what if you could perform inference entirely within memory using In-Memory Computing?

Show Post