Bullish on Automation and Robotics, but not Humanoid Robots | Shahin Farshchi

Shahin Farshchi expresses strong enthusiasm for the robotics sector, driven by macro trends in software and hardware commoditization. He clarifies that while robots are integral to automation, they are not synonymous, emphasizing that specific applications require tailored robotics solutions rather than a one-size-fits-all approach.

Farshchi predicts emerging markets for robotics in areas previously untouched, but he remains skeptical about humanoid robots, viewing them as likely to occupy a niche market. He argues that the push for humanoid robots as universal solutions is more about financial incentives than practicality, as specialized robots are expected to outperform general-purpose ones.

He acknowledges the significant engineering costs associated with developing automation solutions but anticipates a future where individuals can program robots using AI interfaces without extensive technical expertise. Companies like Physical Intelligence are already advancing in automating unstructured tasks through open-source models, making automation tools more accessible to non-experts.

Farshchi compares the current state of robotics to the early 1990s in computing, focusing on physical machines rather than their outputs. He addresses concerns about job displacement due to automation, noting that historical trends show increased automation often correlates with lower unemployment rates. While automation may displace some jobs, it can also create higher-quality positions that enhance worker satisfaction.

He emphasizes the need for workforce education to prepare for the higher-paying jobs that automation will generate. Citing agriculture as an early adopter of automation leading to food abundance, he mentions that industries like automotive and consumer electronics have benefited from increased automation, resulting in safer and more efficient products. Farshchi remains cautious about humanoid robots in domestic settings, predicting that their development will take longer than anticipated due to consumer utility complexities.

Identifying significant opportunities for automation in factory and warehousing environments, he acknowledges the challenges new automation companies face in demonstrating reliability and justifying return on investment. The ability to quantify the value added by automation is a major hurdle for these companies.

Farshchi discusses the multifaceted nature of AI in robotics, highlighting the blend of sensing, perception, planning, and action. He notes the prevalence of large language models (LLMs) but questions their effectiveness in addressing robotics challenges, suggesting that future solutions will require hybrid approaches integrating language models with traditional methods.

He highlights the capabilities of vision-language models (VLMs) and vision-language agents (VLAs) in interpreting scenes and generating actionable plans. Companies that can create robust training sets for robotics will gain a competitive edge in the industry. Farshchi points out the labor-intensive training process for autonomous vehicles, which currently requires hundreds of engineers to ensure safety and reliability.

Distinguishing between LLMs, which primarily function as chatbots, and VLMs and VLAs, designed for scene interpretation and action execution, he mentions Formic as a company actively deploying robots in manufacturing and logistics, aiming to become the largest employer of robots globally. While he expresses personal interest in having robots perform simple tasks at home, he acknowledges that safety concerns hinder widespread adoption.

Looking ahead, Farshchi predicts that VLMs and VLAs will evolve to become faster, more reliable, and easier to train over time, indicating a promising future for automation and robotics in various sectors.