A curated daily news digest covering several AI and tech stories: startups exploring transformer alternatives to make LLMs faster and more efficient; AI academic researchers navigating a shifting landscape; Nvidia securing $500 billion from Wall Street for AI infrastructure; Mark Zuckerberg's open-source AI manifesto alongside Meta's new open-source model; Bernie Sanders calling for a pause on AI development; social media addiction lawsuits proceeding in US courts; Unitree's massively oversubscribed IPO; and China's new rules on emotionally interactive AI.
Table of contents
AI professors are negotiating the new realities of academic researchQuestions this post answers
What are the main limitations of transformer architecture in large language models?
Transformers have two key limitations as LLMs scale: their dense attention mechanism becomes increasingly expensive as the amount of text grows, and they struggle to keep track of large amounts of information simultaneously. These bottlenecks have motivated startups to explore alternative neural network architectures that could make LLMs faster, more efficient, and potentially smarter. Developers building or evaluating LLM infrastructure track emerging architecture shifts on daily.dev.
How much money has Nvidia raised from Wall Street for AI infrastructure?
Nvidia secured $500 billion from Wall Street for AI infrastructure through deals with BlackRock, Goldman Sachs, and four other institutional partners. The arrangement signals that AI infrastructure is being treated as a new asset class by major institutional investors. Engineers and architects following AI compute investment trends find the latest developments on daily.dev.