While Cannes showcases the transformative potential of AI, there's an underlying conversation happening behind the scenes focused on the real financial implications of AI adoption.
The federal government has launched a major initiative to develop a quantum supercomputer, aiming to revolutionize fields such as encryption and drug discovery. This bold investment could drive the next significant shift in technology, creating new opportunities and challenges across industries.
Barclays projects the humanoid robot sector will achieve $200 billion in revenue by 2035, highlighting significant market potential as key technological hurdles are overcome.
OpenAI is deploying artificial intelligence to assist the open source community in identifying and fixing software vulnerabilities more effectively, bolstering security across open source projects.
Menlo Ventures has earned a reputation as a leading AI investor, thanks to a daring $750 million commitment to Anthropic in 2024 that paid off handsomely.
Carolyn Geason-Beissel/MIT SMR | Getty Images
This research analyzes the strategies of leading corporate venture capital (CVC) units from companies like Intel, Cisco, GE, Siemens, and Panasonic, among others. The study reviewed 59 active CVCs from 2017-2024, assessing whether they prioritize strategic benefits that align closely with their parent companies or financial returns from investments. Interviews with experienced CVC managers reveal persistent challenges in balancing these dual objectives.
The authors identify three main CVC investment models: strategic-priority, financial-priority, and hybrid. Strategic CVCs focus on accessing new technologies and business ideas to benefit the parent company, requiring tight integration. Financial CVCs prioritize monetary returns with less parent company involvement. Hybrid CVCs attempt to balance both but often struggle with a lack of clear focus.
The study highlights that CVCs perform best when they clearly define their primary objective—either strategic or financial—and align their operations accordingly. Execution challenges include measuring strategic benefits, maintaining financial discipline, and attracting top investment talent, as compensation and incentives often lag behind independent venture capital firms.
Ultimately, prioritizing strategic investments while maintaining financial rigor tends to offer the most sustainable model. However, CVCs must adapt as parent companies’ goals evolve, and maintaining a delicate balance between close parent collaboration and investment independence remains essential for long-term success.
Not all companies need to create their own cutting-edge AI language models, but most can and should tailor the system that controls these models for their unique needs. Currently, tuning these AI control systems—called harnesses—is largely a manual, intuition-driven process, making it hard to keep up with fast-evolving language models. Researchers at Shanghai Artificial Intelligence Laboratory have introduced "Self-Harness," a novel approach that allows AI agents to systematically refine their own operating rules by analyzing their performance and applying data-driven updates. This self-improving mechanism enables teams to deploy more adaptable and reliable AI agents that evolve alongside their specific models.
The harness includes system prompts, tools, memory, verification protocols, policies, and recovery steps, all of which significantly influence an agent's performance beyond the capabilities of the model itself. Traditional harness engineering suffers from its dependence on manual debugging without systematic feedback, leading to inefficiencies and blind spots. Self-Harness circumvents this by implementing a continuous loop of detecting weaknesses, proposing targeted changes, and validating improvements through rigorous testing.
Applied to a benchmark suite, Self-Harness led to performance gains of 33% to 60% across various language models by making precise changes that address recurring execution failures. Examples include adding rules to prevent infinite loops, enforcing command retry limits, and preserving environment settings. While powerful, this approach requires substantial computational resources and depends on accurate evaluation mechanisms, making it most suitable for environments where failures are measurable and consequences of trial and error are manageable. It is less appropriate for contexts demanding subjective, delayed, or high-risk evaluations.
Rather than eliminating human engineers, Self-Harness shifts their role toward architecting feedback systems that enable continuous AI improvement, highlighting a broader transition in AI development from manual tweaking to designing robust, data-driven optimization frameworks.
Sakana, an enterprise-focused AI startup, launched Fugu, a cutting-edge multi-agent orchestration system designed to deliver top-tier AI performance through a single, OpenAI-compatible API. Unlike traditional monolithic models, Fugu dynamically routes queries through a diverse pool of specialized AI agents, ensuring resilience against vendor lock-in and geopolitical restrictions. Co-founded by former Google Brain lead David Ha, Fugu matches and sometimes exceeds the performance of restricted models like Anthropic's Claude Fable 5 on benchmarks, while offering scalability and operational flexibility not found in single-provider AI systems. It supports two variants: Fugu for everyday tasks and Fugu Ultra for complex, high-stakes workloads such as AI research and cybersecurity. Pricing varies with usage levels, with Fugu Ultra having a premium, fixed rate. While Fugu excels in managing multi-step AI tasks via its collective intelligence model, it occasionally falls short of the highest performing specialized models in specific domains. Sakana emphasizes the proprietary nature of their model coordination technology, offering enterprises control over data privacy and model selection. Overall, Fugu represents a strategic shift in enterprise AI, prioritizing model orchestration and vendor redundancy over monolithic scale, catering to developers, enterprises, and governments wary of sudden access restrictions to high-end AI models.
After nearly two decades of concentrating its efforts in Europe, Seedcamp, an early-stage investment firm, announced the successful raise of $320 million for its newest fund. This capital injection will support the firm's strategic move to increase its footprint in the United States, marking a significant expansion beyond its traditional European market focus.
Will Cathcart is transitioning to a new role within Meta, paving the way for Kunal Shah, founder of Indian fintech firm CRED, to step down as CEO and become the new head of WhatsApp. Alongside this leadership shift, Meta has announced a $900 million investment in an Indian startup, signaling a deepened focus on the region's tech ecosystem.
