# Sep 6, 2024

## Daily Updates

### AlphaProteo generates novel proteins for biology and health research
- **AlphaProteo**, a new AI system by Google DeepMind, designs novel proteins with high binding affinities, potentially revolutionizing drug design and disease understanding.

### Reflection 70B, the top open-source model
- **Reflection 70B**, an **open-source model**, introduces a novel **Reflection-Tuning technique** for self-correction in LLMs, setting a new standard in AI development.

### Show HN: Infinity – Realistic AI characters that can speak
- **Infinity AI** has developed a **video diffusion transformer model** that generates **realistic AI characters** capable of speaking, driven by **audio input**, marking a first in the field.

### Launch HN: Maitai (YC S24) – Self-Optimizing LLM Platform
- **Maitai introduces a self-optimizing LLM platform** that enhances request routing, autocorrects errors, and fine-tunes application-specific models for incremental improvements, demonstrated through a [demo video](https://www.loom.com/share/a2cd9192359840cab5274ccba399bd87?sid=7097fd84-ea85-42cd-9616-84abc1087a56) and an interactive game at [maitaistreasure.com](https://maitaistreasure.com/).

### The Memory Wall: Past, Present, and Future of DRAM
- **DRAM scaling has significantly slowed**, with density increases now only doubling every decade compared to the historical rate of doubling every 18 months, highlighting a stark departure from the rapid advancements seen in the past.

### Effects of Gen AI on High Skilled Work: Experiments with Software Developers
- The study found a **26.08% increase in productivity** among software developers using GitHub Copilot, as evidenced by the completion of more tasks, based on data from three field experiments involving 4,867 participants.

### What if self-attention isn’t the end-all be-all?
- **Masked Mixers** propose an **alternative to self-attention mechanisms** in transformers, addressing **information loss** issues.

### Is exploration the key to unlocking better recommender systems?
- **Google DeepMind's research** suggests that **exploration** in recommendation systems can **significantly enhance long-term user experience** by diversifying content.

### Retrieval-Augmented Generation vs Long-context LLM
- **Recent discussions** highlight doubts about **Long-context Language Models (LC-LLMs)** fully replacing **Retrieval-Augmented Generation (RAG)** due to challenges in managing **vast data contexts** and **computational costs**.

### Debate on Graph: a Flexible and Reliable Reasoning Framework for Large Language Models
- **Debate on Graph (DoG)** introduces an **iterative interactive framework** for Large Language Models (LLMs) to enhance reasoning by focusing on relevant subgraphs and employing a multi-role debate team to simplify complex questions.

### Planning In Natural Language Improves LLM Search For Code Generation
- **PLANSEARCH**, a novel search algorithm, **enhances code generation** by leveraging **natural language planning**, leading to a more diverse set of potential solutions.

### A Different Level Text Protection Mechanism With Differential Privacy
- The article introduces a **novel text protection method** that leverages the **BERT pre-training model** to **extract words of varying importance**, enhancing data privacy without compromising text utility.

### CogniDual Framework: Self-Training Large Language Models within a Dual-System Theoretical Framework for Improving Cognitive Tasks
- The **CogniDual Framework for LLMs (CFLLMs)** aims to **emulate human cognitive development** in large language models (LLMs) by transitioning from deliberate to intuitive processing through self-training.

### Sketch: A Toolkit for Streamlining LLM Operations
- **Sketch**, an innovative toolkit, aims to **streamline LLM operations** by addressing the challenge of controlling and harnessing the model's outputs across diverse fields.