#161 - Claude 3 beats GPT-4, Stability CEO resigns, DBRX, TacticAI, UN resolution on AI
Mar 31, 2024
auto_awesome
In the latest discussion, Claude-3 surpasses GPT-4 in user preference, signaling a shift in AI development. The stability AI CEO resigns to focus on decentralized solutions, while Amazon's hefty investment in Anthropic underscores the booming AI sector. Samsung leads the smartphone AI race, challenging Apple. The podcast also highlights Intel's impressive Gaudi 2 performance in comparison to NVIDIA's H100. Plus, a UN resolution raises ethical questions around AI and human rights, emphasizing the need for responsible deployment in the tech landscape.
Claude-3 Haiku matches GPT-4 level preference, showcasing user-driven AI advancement.
Stability AI CEO resigns for decentralized AI pursuit, signaling industry shifts.
Intel Gaudi 2 competes with NVIDIA H100 in MLPerf benchmarks, shaping AI hardware landscape.
Tactic AI models optimize football corner kicks, blending deep learning with sports expertise.
Deep dives
Databricks Unveils D.B. Rax, a State-of-the-Art Language Model Outperforming Predecessors
D.B. Rax, a new general purpose language model by Databricks, surpasses GPT 3.5 and competes with Gemini One Pro, trained on 12 trillion tokens of text and code data, offering impressive coding performance and speed.
Celestial AI Secures a Massive $175M Series C Funding for AI-Powered Robotics
Celestial AI, known for its photonic fabric platform improving robot manipulation, lands a significant $175 million in Series C financing, enhancing capabilities for faster and more energy-efficient AI processing.
Apollo, the Texas-Based Robot, Joins the Workforce, Aiding in Assembly and Inspection Tasks
The robot 'Apollo' from Texas startup Apptronic joins a manufacturing plant in Hungary, skilled in carrying up to 55 pounds and assisting in inspecting assembly kits and delivering parts to workers, showcasing a blend of robotic functionalities.
Play US Collaborates with DeepMind to Develop Tactic AI for Livable FC
DeepMind teams up with Livable FC to create Tactic AI models using geometric deep learning to optimize corner kick setups in football, achieving preferred outcomes by human expert wagerers nearly 90% of the time, demonstrating deep specialization in sports-specific contexts.
Study on Persuasion by Large Language Models Through Debates
Participants in a randomized controlled trial engaged in debates with human opponents and large language models. The study found that participants debating GPT-4 with access to personal information showed increased agreement with opponents. The results highlighted the potential persuasiveness of AI, especially when given access to demographic data.
Insights into Transformer Computation Process
Research delved into transformer language models' computation process, revealing that crucial information is encoded in a neural network's early layers. Linear transformations of representations at these layers can predict outcomes accurately. The study emphasized the significance of understanding internal workings of transformers for interpretability and control.
US Imposes Sanctions on Chinese Chip Networks
The US announced sanctions on Chinese semiconductor firms associated with Huawei, aiming to limit their access to cutting-edge technology. These measures signify efforts to counter China's advancements in chip design and manufacturing, especially with implications for AI capabilities and global technological competition.
Our 161st episode with a summary and discussion of last week's big AI news!
Check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube.