Generative AI is moving beyond lab experiments and is beginning to power critical business workflows. This shift has fundamentally transformed the security landscape. Unlike traditional software, Large Language Models (LLMs) and agentic systems introduce vulnerabilities more akin to human fallibility than traditional code exploits. Since LLMs generate responses probabilistically rather than deterministically, carefully crafted inputsContinue reading “Securing Generative AI: Beyond Traditional Playbooks”
Author Archives: Ben Lorica
Is Your AI Secure? New Threats & How to Mitigate Them
Subscribe • Previous Issues Securing Generative AI: Beyond Traditional Playbooks Generative AI is moving beyond lab experiments and is beginning to power critical business workflows. This shift has fundamentally transformed the security landscape. Unlike traditional software, Large Language Models (LLMs) and agentic systems introduce vulnerabilities more akin to human fallibility than traditional code exploits. Since LLMs generateContinue reading “Is Your AI Secure? New Threats & How to Mitigate Them”
US vs. China: Who Wins the Critical AI Diffusion Battle?
When comparing the United States and China in artificial intelligence, the spotlight is on the development of foundation models. At first glance, America maintains a comfortable numerical lead, producing 40 notable models in 2024 compared to China’s 15. However, looking deeper reveals a more complex story: China’s best models, particularly from DeepSeek, Alibaba (Qwen), andContinue reading “US vs. China: Who Wins the Critical AI Diffusion Battle?”
The Real AI Race: It’s About Diffusion
Subscribe • Previous Issues US vs. China: Who Wins the Critical AI Diffusion Battle? When comparing the United States and China in artificial intelligence, the spotlight is on the development of foundation models. At first glance, America maintains a comfortable numerical lead, producing 40 notable models in 2024 compared to China’s 15. However, looking deeper reveals aContinue reading “The Real AI Race: It’s About Diffusion”
Navigating Huawei’s AI Hardware: Bright Spots, Backdrop, Barriers
In response to Western technology restrictions, Huawei has advanced its artificial intelligence hardware capabilities, improving manufacturing yields for its Ascend 910C chips while introducing the next-generation Ascend 920. Rather than pursuing peak per-chip performance, where it faces external constraints, the firm employs a scale-out strategy embodied by its CloudMatrix 384 system. This architecture aggregates 384Continue reading “Navigating Huawei’s AI Hardware: Bright Spots, Backdrop, Barriers”
Agents at Work: Navigating Promise, Reality, and Risks
Agents are top of mind for people working in AI. Still, when I talk to professionals building AI applications, many express frustration, highlighting the gap between the intense interest in agents and their relatively limited presence in live enterprise environments. Part of this skepticism is justified—as evidenced by the systemic failure modes we recently exploredContinue reading “Agents at Work: Navigating Promise, Reality, and Risks”
Real-World Lessons from Agentic AI Deployments
Subscribe • Previous Issues Agents at Work: Navigating Promise, Reality, and Risks Agents are top of mind for people working in AI. Still, when I talk to professionals building AI applications, many express frustration, highlighting the gap between the intense interest in agents and their relatively limited presence in live enterprise environments. Part of this skepticism isContinue reading “Real-World Lessons from Agentic AI Deployments”
Beyond the Hype: The Reality Gap in Multi-Agent Systems
The allure of multi-agent systems (MAS), where teams of LLM-based agents collaborate, is undeniable for tackling complex tasks. The theoretical benefits seem clear: breaking down problems, parallelizing work, and leveraging specialized skills promise more sophisticated AI solutions than single agents can deliver. Yet as teams building these systems are discovering, translating this promise into reliableContinue reading “Beyond the Hype: The Reality Gap in Multi-Agent Systems”
Google’s AI Revival
Support our work by subscribing to our newsletter🎁
Why Your Multi-Agent AI Keeps Failing
Subscribe • Previous Issues Beyond the Hype: The Reality Gap in Multi-Agent Systems The allure of multi-agent systems (MAS), where teams of LLM-based agents collaborate, is undeniable for tackling complex tasks. The theoretical benefits seem clear: breaking down problems, parallelizing work, and leveraging specialized skills promise more sophisticated AI solutions than single agents can deliver. Yet asContinue reading “Why Your Multi-Agent AI Keeps Failing”