Watch this video without interruptions
In a recent interview with The Jerusalem Post, Ely Razin, a seasoned fintech leader, discussed various aspects of AI, the challenges in deploying AI in real-world applications, and the role of investors in deep-tech ventures.
Razin, who has a track record of multiple startup exits and now serves as a chief strategic investor at OurCrowd, shared insights into the evolving field of physical AI.
Razin emphasized the differences between traditional AI housed in large data centers and the emerging field of edge AI, which involves smaller, localized devices such as robots, cameras, and cars. These edge devices must optimize power consumption and minimize heat generation, making physical AI uniquely challenging. He explained that physical AI isn't just about collecting data; it also requires significant advancements in how machines interact with the real world, a challenge that companies are beginning to tackle.
In discussing AI's scalability, Razin noted that while AI chips are crucial, data scarcity remains a key barrier for many companies. He highlighted synthetic data as a solution to this problem, with companies like Meta (Facebook) investing in creating artificial datasets for training AI systems.
Additionally, Razin noted that once trained, AI models enter the "inference" phase, where they perform real-world tasks, underscoring the importance of ensuring they produce reliable outputs.
Addressing concerns in sectors like finance and healthcare, where data security and privacy are paramount, Razin discussed the trend of companies opting for on-premises data processing over cloud solutions to maintain data control and reduce exposure to risks associated with cloud-based storage.
He also touched on the geopolitical competition surrounding AI, with countries like the US and China investing heavily in the infrastructure required for AI and quantum developments.
Lastly, Razin discussed the potential of quantum computing and its impact on AI. He noted that while quantum computing is still evolving, it will eventually complement AI, enhancing computational power and enabling breakthroughs.