**This Week's Top Stories in AI**
1. **DeepMind Unveils Human-Teaching AI System**
Google DeepMind has introduced an innovative AI agent system capable of learning complex tasks directly from human instructors. In a groundbreaking paper published in *Nature*, researchers introduce a method termed cultural transmission, which enables the AI model to learn skills without relying on pre-collected human data. This novel approach closely mirrors the way individuals learn through observation, akin to following a video tutorial. Through a unique few-shot imitation process demonstrated in a simulated environment called GoalCycle3D, the AI agent successfully mimicked actions from a human demonstrator, showing impressive recall of tasks long after the demonstration ended. DeepMind asserts that this technique sets the stage for integrating cultural evolution into the development of artificial general intelligence, significantly advancing the field.
2. **Preparing for the AI Workforce Revolution**
In an insightful column by Seth Dobrin, founder and CEO of Qantm AI, the discussion centers on how organizations and employees can proactively leverage AI advancements. With the rapid evolution of technology, it’s crucial for today’s workforce to reskill and adapt swiftly to an AI-driven economy. Dobrin highlights the urgent need for educational systems, corporate training programs, and government policies to evolve in tandem with these changes. The impact of generative AI spans all sectors, affecting both manual labor and white-collar roles. A coordinated effort across various stakeholders is essential to prepare for this transformative period and minimize workforce disruption.
3. **Google Launches Gemini: A Powerful New Language Model**
Google has officially launched its much-anticipated next-generation large language model, Gemini, touted as the largest and most capable model in its lineup. After a period of anticipation, Google released detailed insights about Gemini, which comes in three configurations: Ultra, Pro, and Nano, allowing it to be utilized across devices from mobile phones to data centers. Gemini Pro is now powering Google's Bard chatbot, enhancing its reasoning and understanding capabilities. In early 2024, Google plans to introduce Bard Advanced, with additional functionalities. Starting December 13, developers and enterprise users will gain access to Gemini Pro via the Gemini API on Vertex AI and Google AI Studio.
4. **Meta’s Audiobox: Transforming Text into Unique Sounds**
Meta has launched Audiobox, its latest AI model designed for audio generation that translates text into sound. Unlike its predecessor, Voicebox, Audiobox allows users to describe the audio they wish to create using natural language prompts. For instance, typing ‘a beaver munching on a slice of pineapple’ generates a unique soundscape based on that description. Additionally, the model can synthesize audio by combining text prompts with voice inputs, offering users enhanced control over the generated audio's style and sound effects—a feature not available in earlier models. Experience some audio samples on Meta’s research website.
5. **OpenAI's Strategic Chip Purchase from Rain AI**
OpenAI has entered into a significant $51 million agreement to acquire AI chips from Rain AI, a startup co-founded by CEO Sam Altman. This collaboration, rooted in advanced neuromorphic processing technology, aims to replicate how the human brain processes information. OpenAI's previous engagements with Rain date back to 2019, highlighting a longstanding partnership in innovation. Despite recent internal controversies surrounding Altman, the deal is poised to enhance OpenAI's capabilities, with Rain expecting to deliver its first hardware batch by October 2024.
Ranking
