• Posts
  • Gemini breaks new ground with a faster model, longer context, AI agents and more

Gemini breaks new ground with a faster model, longer context, AI agents and more

Also: OpenAI’s Co-Founder and Chief Scientist Ilya Sutskever Is Leaving the Company


In today's major tech developments, Google has announced significant enhancements across its AI platforms, notably with its Gemini 1.5 models and the new Trillium AI chip, which boasts a 67% increase in energy efficiency. Meanwhile, OpenAI sees a pivotal shift as co-founder Ilya Sutskever departs to pursue new endeavors, passing the chief scientist baton to Jakub Pachocki. In legislative circles, a Senate study has proposed a substantial annual investment of $32 billion to boost AI infrastructure and national security in the U.S., highlighting the growing importance of AI in strategic national domains. Conversely, Microsoft faces challenges reconciling its aggressive AI expansion with its climate commitments, as AI-driven energy demands strain its environmental goals. These updates underscore a dynamic landscape where technological advancements in AI are increasingly intersecting with broader socio-economic and environmental considerations.


  • 🆕 Gemini breaks new ground with a faster model, longer context, AI agents and more

  • ⚡️ Trillium: Google unveils its most potent AI chip with 67% energy efficiency

  • 👀 OpenAI’s Co-Founder and Chief Scientist Ilya Sutskever Is Leaving the Company

  • 💵 Senate study proposes ‘at least’ $32B yearly for AI programs

  • 🌧️ Microsoft’s AI obsession is jeopardizing its climate ambitions

Google's Gemini updates feature several advancements, including the Gemini 1.5 Flash model, which is optimized for speed and efficiency, and designed for high-volume, high-frequency tasks. This model, lighter than the Gemini 1.5 Pro, still retains a long context window of up to 1 million tokens, and can efficiently handle tasks such as summarization, chat applications, and data extraction from complex documents. The Gemini 1.5 Pro has also been enhanced with improved code generation, logical reasoning, and multimedia capabilities, supporting extended context windows up to 2 million tokens. Additionally, Google has introduced Gemma 2, the next generation in its family of open models, offering new architectural improvements and performance efficiencies. Another significant development is Project Astra, which aims to create more responsive and conversational AI agents, able to understand and interact in real-time using multimodal inputs. These updates collectively push the boundaries of AI applications, improving both functionality and accessibility for developers and users alike.

Google recently introduced the Trillium chip, its latest advancement in AI-specific hardware, which stands out for its 67% energy efficiency improvement over previous models. The sixth-generation Trillium Tensor Processing Units (TPUs) offer a significant 4.7X boost in peak compute performance, accompanied by enhancements in High-Bandwidth Memory (HBM) capacity and bandwidth, as well as Interchip Interconnect (ICI) bandwidth. Notably, Trillium TPUs feature third-generation SparseCore technology, which is crucial for processing complex embeddings typical in ranking and recommendation engines, allowing for the training of foundation models with reduced latency and costs. The chip's scalable architecture can connect up to tens of thousands of chips across multiple pods, promoting large-scale AI computations in a multi-petabit-per-second datacenter network. This technology is available exclusively via Google's cloud services, reflecting Google's strategy to enhance its infrastructure to support more demanding AI applications.

Ilya Sutskever, co-founder and chief scientist at OpenAI, has announced his departure from the company. Sutskever has been instrumental in shaping OpenAI since its inception, playing a critical role in its strategic direction and research efforts. His departure marks the end of a significant chapter in the company’s history, during which he contributed to the development of groundbreaking AI technologies and navigated complex challenges, including ethical concerns about AI’s impact on society. His exit follows a period of internal discord highlighted by his brief role in the attempted ouster of CEO Sam Altman last year, which he later regretted. Sutskever plans to focus on a new, personally meaningful project, while OpenAI has appointed Jakub Pachocki, a key figure in the development of the GPT-4 model, as the new chief scientist.

A recent Senate study recommends an annual allocation of at least $32 billion for AI programs in the United States, focusing on infrastructure, innovation, and national security enhancements. This initiative is outlined in a report from a bipartisan Senate working group spearheaded by Senator Chuck Schumer. Although this roadmap is not yet formalized into legislation, it underscores the significant investment required to maintain the U.S.'s competitive edge in global AI technology. The proposed funding aims to support various sectors including AI hardware and software advancements, national AI research resources, AI readiness in cybersecurity for elections, and the modernization of government services using AI. This extensive financial commitment reflects the growing recognition of AI's pivotal role in advancing national interests across multiple dimensions.

Microsoft's ambitious AI developments are increasingly at odds with its climate commitments, posing significant challenges to its goal of becoming carbon negative by 2030. Since 2020, when Microsoft pledged to dramatically reduce its carbon footprint and invest in carbon capture technologies, its greenhouse gas emissions have surged by 30 percent due to its intensified focus on AI. The energy demands of AI technologies, especially those needed for training and running large AI models, are substantial and growing. Despite making the largest renewable energy purchase in history to power these AI operations, the scale of AI's energy consumption is making it difficult for Microsoft to balance its technological advancements with its environmental responsibilities. This scenario underscores a growing conflict within the tech industry—pursuing cutting-edge AI while striving to adhere to ambitious climate targets.

🛠️ AI tools updates

Google's Project Astra, announced during Google I/O 2024, represents a significant step towards creating real-time, multimodal AI assistants designed to enhance everyday user experiences. Engineered by Google DeepMind, Project Astra operates on the Gemini 1.5 Pro and utilizes 1 million tokens to efficiently address complex user needs through enhanced long-context capabilities. The AI can interact via a smartphone camera to recognize and recall objects and text, responding intelligently to user queries. This technology can even remember and locate objects like misplaced glasses. Unlike previous models that primarily served as information libraries, Project Astra personalizes interactions, aiming to resolve specific user challenges through contextual understanding and responsiveness, marking a pivotal development in AI as it moves towards more personalized and context-aware applications.

💵 Venture Capital updates

Weka, a company specializing in data management for AI applications, has secured $140 million in a recent Series E funding round. This investment round was led by Valor Equity Partners and included contributions from several prominent technology firms such as Nvidia and Qualcomm Ventures. Weka’s platform addresses the critical challenges of data disorganization and inefficiency, which are common obstacles in developing AI technologies. Their solution facilitates the efficient handling and processing of diverse data types and sources, thereby enhancing the capability of AI systems to operate with higher performance and less downtime. Weka's system is particularly noted for its ability to speed up AI model training by optimizing data movement across storage locations, which is a significant advantage given the increasing complexity and data demands of modern AI workloads. This funding positions Weka to expand its technology and market reach, amidst a burgeoning industry where effective data management is crucial for advancing AI development.

🫡 Meme of the day

⭐️ Generative AI image of the day