Welcome back to AI Hungry, where the latest developments in artificial intelligence are always on the menu. In this update, we delve into Databricks' unveiling of DBRX, a new open source AI model that's stirring the competitive pot. Additionally, we'll explore the significant shift at Stability AI, as CEO Emad Mostaque steps down to chart a new course in decentralized AI.
These stories not only reflect individual company strategies but also signal broader industry trends. From the pursuit of efficiency and cost-effectiveness in AI models to the leadership changes that pivot towards decentralization, there's much to unpack in the evolving landscape of artificial intelligence.
Databricks Unveils DBRX: A New Open Source AI Model Challenging Industry Giants
Databricks has introduced DBRX, an open source artificial intelligence model boasting 132 billion parameters, surpassing the performance of other open source models in language, programming, and math benchmarks. While not as powerful as OpenAI's GPT-4, DBRX is presented as a more efficient and cost-effective alternative to GPT-3.5, aiming to push the boundaries of open source AI capabilities.
DBRX's innovative 'mixture-of-experts' architecture allows it to operate with fewer active parameters, enhancing performance and reducing costs. By open sourcing DBRX, Databricks not only advances its AI research leadership but also supports its core business of custom AI model development, offering clients advanced AI benefits while ensuring data security.
Stability AI CEO Emad Mostaque Steps Down to Pursue Decentralized AI
Emad Mostaque, the founder and CEO of Stability AI, has stepped down from his role and the company's board. This significant change in leadership comes as Stability AI faces financial hurdles and staff departures. Mostaque, who remains the majority shareholder, has expressed a desire to explore decentralized AI, criticizing the current centralized models of AI governance. In the interim, COO Shan Shan Wong and CTO Christian Laforte will serve as joint CEOs.
The company, known for creating the popular Stable Diffusion image software, has been struggling financially, with a reported monthly burn rate of $8 million and unsuccessful funding efforts at a $4 billion valuation. Mostaque's resignation reflects a broader industry trend of reevaluating the direction of AI development, with a focus on transparency, distributed governance, and the potential of open models in the face of rapid technological advancements.
🎨 Adobe Firefly's New Features Revolutionize Image Creation at Adobe Summit 2024. Adobe unveils new Firefly features, including Structure Reference for image layout replication and Custom Models for brand-specific design training. Firefly Services to automate content creation, with Microsoft integrating these tools into its applications. (Link)
💼 Amazon Deepens AI Investment with Additional $2.75 Billion in Anthropic. Amazon boosts its stake in AI startup Anthropic with an extra $2.75 billion investment, totaling $4 billion. The partnership includes AWS as Anthropic's primary cloud provider and aims to advance generative AI solutions in regulated industries. (Link)
💡 Foundry Launches to Democratize AI Compute with $80M Funding. Former DeepMind scientists have created Foundry, a cloud platform designed to simplify access to AI training hardware. With $80 million in funding, Foundry aims to challenge major cloud providers by offering scalable AI hardware solutions. (Link)
🔒 California Sets Strict Guidelines for Generative AI in State Government. California has established new rules for state agencies using generative AI, mandating audits, risk assessments, and expert consultations before procurement. The guidelines aim to ensure transparency and prevent harm to citizens. (Link)
🔬 Revolutionary AI Designs Antibodies to Combat Superbugs. Researchers at Stanford and McMaster Universities have developed SyntheMol, an AI that can create billions of antibiotic molecules to fight drug-resistant bacteria, including the dangerous A. baumannii. (Link)
🤖 MIT Engineers Empower Robots with Common Sense Using Language Models. MIT engineers have created a new framework, Glide, that uses language models to give robots common sense for tasks. This innovation automates the learning process, allowing robots to understand their environment and actions needed to complete tasks. (Link)
