ConnectingAI #103: Broadcom boosts integrated data center optics and more
Bits and bites of how hardware enables the future of AI, the latest technological advancements and my favorite news in AI this week.
Top ConnectingAI articles of the week
1- Broadcom boosts integrated data center optics: Broadcom and Corning have partnered to develop co-packaged optics solutions, integrating optical and silicon components to enhance data center performance. This collaboration aims to reduce power consumption and increase bandwidth, marking a significant advancement in data center infrastructure.
2- AI Open-Source Projects That Should Be on Your Radar: Open-source initiatives are accelerating AI development by promoting transparency, modularity, and vendor neutrality. These projects enable rapid innovation and flexibility, allowing enterprises to adapt AI solutions to their specific needs without being tied to proprietary systems.
3- AI agents and efficient IT: Chris Wolf of Broadcom highlights the role of AI agents in improving IT efficiency, citing use cases like contact center support and coding assistants that offer significant productivity gains. He emphasizes the importance of privacy, control, and flexibility in deploying AI models across various environments, including private clouds and edge sites.
4- AI and cybersecurity today.: AI is transforming cybersecurity by enabling faster, more sophisticated attacks, such as AI-generated phishing and deepfakes. Conversely, AI also strengthens defense mechanisms through real-time threat detection and automated response, making it both a challenge and a solution in cybersecurity.
5- Relying on file storage heritage, Box pivots to AI: Box is leveraging its expertise in file storage to pivot towards AI-driven content management, aiming to enhance how users interact with data. CEO Aaron Levie envisions AI agents fundamentally changing the value of content and the way people work with information.
6- Startup working to speed up chip development: Cognichip has raised $33 million in funding to expedite the development of its AI chips, focusing on improving performance and energy efficiency. The investment will support the company's efforts to meet the growing demand for advanced AI hardware solutions.
AI piqued my interest
1- AI tool predicts biological age by looking at a face: Researchers at Harvard and Mass General Brigham have developed FaceAge, an AI tool that estimates a person's biological age by analyzing facial photographs. The tool, trained on over 58,000 images, can aid oncologists in tailoring treatments by providing insights into a patient's physiological health beyond chronological age.
2- MIT disavows doctoral student paper on AI’s productivity benefits: MIT has disavowed a doctoral student's paper claiming AI tools increased productivity in a materials science lab, citing concerns over data integrity and validity. The university has requested the withdrawal of the paper from publication platforms, emphasizing the importance of research reliability.
3- AI models can't tell time or read a calendar, study reveals: A study presented at the 2025 International Conference on Learning Representations reveals that AI models have difficulty interpreting analog clocks and calendars. The findings highlight limitations in AI's ability to process temporal and spatial information, posing challenges for time-sensitive applications.
4- AI Will Soon Power In-Car Android Auto: Google's Gemini AI assistant is being integrated into Android Auto, enabling drivers to interact with their vehicles using natural language for tasks like messaging and navigation. The multimodal AI aims to enhance hands-free functionality and personalization in vehicles from manufacturers such as Lincoln, Renault, and Honda.
5- AI Chatbots Mirror a Human Brain Disorder: Researchers at the University of Tokyo have found that LLMs produce fluent yet sometimes incoherent responses, mirroring language patterns seen in Wernicke’s aphasia patients. The study suggests that understanding these parallels could improve both AI language processing and clinical approaches to aphasia.
6- Meet AlphaEvolve, AI that writes its own code—and saved millions in computing costs: Google's AlphaEvolve AI has demonstrated the ability to autonomously write and optimize its own code, leading to significant reductions in computing costs. By streamlining code development, AlphaEvolve enhances efficiency and performance in AI-driven applications.