Are AI politicians next?

Meta AI's New LLM Is Going Mobile

Meta AI just unveiled MobileLLM, a language model designed specifically for mobile devices. With fewer than 1 billion parameters—compared to the trillion-parameter behemoth that is GPT-4—MobileLLM challenges the belief that bigger is always better in AI.

Why mobile matters: Most of us interact with technology through our smartphones on a daily basis. MobileLLM promises to bring advanced AI capabilities right to these devices, making sophisticated AI that can handle complex queries more accessible and efficient for everyday use—on our phones, without the need for massive cloud computing (sign me up).

Key features:

• Model design: The new LLM focuses on creating a deep, efficient model. It uses embedding sharing to reuse information across the network and grouped-query attention to better channel complex patterns from the input. Bonus? The focus on efficiency reduces energy needs…which in turn makes scaling more realistic.

• Performance: MobileLLM outperforms larger models like LLaMA-2 in specific tasks (for example: voice assistants and real-time language translation) while maintaining a compact size.

What’s next? Mostly research and development. Meta is open-sourcing the pre-training code, which allows other researchers to build on their work and potentially accelerate the development of efficient mobile AI models.

Anthropic's Prompt Playground for Claude

Anthropic just announced its new prompt playground for Claude, designed to make prompt engineering easier and more efficient for developers. This tool aims to drastically reduce the time and expertise needed to fine-tune AI performance. Here’s what you should know →

What is it: Anthropic’s developer console is like an experimental playground for developers, designed to attract businesses that want to build products with Claude. 

Some key features?

• Built-in prompt generator: Constructs detailed prompts from brief descriptions using Anthropic’s advanced prompt engineering techniques

• Evaluate tab: Enables developers to test prompt effectiveness in various scenarios

Who it’s for:

• Developers and AI engineers: Ideal for those building AI applications with Claude—whether novice or experienced

• Businesses and enterprises: Helps create more effective and specialized AI applications, reducing development time and enhancing performance

Unlock Creativity with Generative AI

Explore the world of Generative AI with Coursera's microlearning course, Introduction to Generative AI.

In this flexible, beginner-level course, you’ll learn:

• What Generative AI actually is and how you can use it

• How Generative AI works and its applications

• Generative AI model types and how they differ from traditional ML methods

• How to use Google Tools to develop your own Gen AI apps

3 Takeaways from Writer’s New Tech

Writer, the full-stack generative AI platform for enterprises, just rolled out some major updates this week.

But first, what’s Writer? Writer helps big businesses like Accenture and L’Oreal build and integrate AI-powered applications into their processes. The platform combines proprietary LLMs, a knowledge graph for sophisticated retrieval-augmented generation (RAG), and AI guardrails to ensure accurate, compliant, and on-brand outputs.

Here’s three standout features from this week’s update:

1. 10 million-word capacity: Writer’s chat apps can now crunch through up to 10 million words of company-specific data. This means businesses can parse through lengthy documents, research papers, and extensive documentation. (Better data → better insights)

2. AI “thought process” transparency: Ever wonder how AI makes its decisions? With the new "thought process" feature, Writer users can peek behind the curtain of AI decision-making. It breaks down queries and shows exactly which data sources were referenced, building trust by mapping out how it connects the dots.

3. Dedicated modes: Writer has rolled out new modes for tasks like document analysis, managing data structures, and working with knowledge graphs. But the real standout is the Knowledge Graph, a graph-based RAG system that tightly integrates AI with your company’s data. And in practice? Imagine support bots providing precise answers from detailed product documentation.

• Samsung unveils the Galaxy Ring, a smart ring that brings AI to your fingertips (literally).

• Microsoft and Apple abandon plans to join OpenAI Board.

• OpenAI and Los Alamos National Laboratory team up to advance bioscience research.

• Stability AI unleashes new stable assistant features.

• This robot explores around Google DeepMind offices using Gemini.

• OpenAI’s China ban ignites a frenzy for domestic AI solutions.