Chief AI Officer Blog - Generative AI, A New Paradigm in Computing: Part 2

Author:

Sulabh Soral

In "Generative AI, A New Paradigm in Computing: Part 1,” we explored how advancements in AI are redefining the boundaries of possibility and are democratising the power of computing.ⁱ Building on these themes, in Part 2 we will delve into the exciting convergence of AI and robotics, looking at how the virtual impact of AI is being translated into tangible, physical impact.

As we touched upon in Part 1, embedding GenAI and Machine Learning into robotics is revolutionising how machines interact with both the digital and physical realms. This fusion is poised to reshape industries and unlock unprecedented capabilities, but not without presenting ethical considerations and challenges that we must carefully navigate.

Reimagining Robotics with Generative AI

Once confined to repetitive, pre-programmed tasks, robots are now being transformed into intelligent, adaptive systems capable of complex reasoning and decision-making. Modern robotics is witnessing a shift from explicit programming to intuitive interaction. Traditionally, robots operated as rudimentary ’dumb’ machines, relying on hydraulic-based joints and extensive pre-programming to perform single-purpose tasks in controlled environments like factories.ⁱⁱ As a result, robots have often been expensive, slow to train and restricted in their capabilities. Today, advancements in electric motors, natural language processing, and AI models like Large Language Models (LLMs) are empowering robots to process instructions, adapt to their environment, and even simulate internal reasoning—often referred to as ‘chain of thought reasoning’.

The integration of GenAI has enabled robots to process vast amounts of multimodal data (text, images, video, and sensor inputs), translate it into actionable tasks and learn from themselves. In particular, Vision-Language-Action Models (VLAMs) help robots interpret their surroundings with context-aware precision and common sense, such as the ability to handle fragile objects delicately.ⁱⁱⁱ Augmented computer vision and spatial reasoning capabilities have allowed robots to gain greater autonomy while navigating varied environments and a better understanding of physical, real-world challenges. There have also been advances in hardware and infrastructure, and the evolution of robotics is supported by innovations in chipsets, such as NVIDIA Jetson Thor and AI-specific system-on-a-chip (SoCs) from Intel and Qualcomm, which enhance computational power and real-time decision-making capabilities.^iv These advancements reduce reliance on cloud systems, improving security and latency.

Despite significant technological strides, robotics in many industries remains in its nascent stages. Factory robots are typically designed for repetitive tasks, such as assembly line operations or material handling, with minimal adaptability to changing conditions. Moreover, programming and integration costs account for more than half of deployment expenses, creating barriers to adoption for smaller companies. However, the introduction of AI-powered robotics is beginning to address these limitations, offering cheaper, more flexible, and easily reprogrammable solutions. The decreasing cost of robotics, galvanised by AI innovation and declining hardware expenses, is making robotics more accessible to smaller enterprises and even consumers.

According to Statista, the global average cost of an industrial robot has dropped from $46,000 in 2010, to $27,000 in 2017, with projections forecasting the price to be $10,856 by 2025.^v For reference, Tesla’s humanoid robot, Optimus, is estimated to be priced between $20,000 and $30,000.^vi

The growing market for intelligent robots

The convergence of AI and robotics is creating an era where physical and digital workflows are seamlessly integrated, and AI-driven robots have many cross-industry use cases. In warehousing, robots manage millions of packages daily, such as in Alibaba’s Cainiao logistics network, which ensures efficient operations during peak demand.^vii Construction robots like Hadrian X, developed by FBR (Fastbrick Robotics), streamline the brick-laying process, accelerating timelines and minimising waste while maintaining precision.^viii In the realm of scientific research, the start-up Tetsuwan Scientific has developed an AI robot scientist that can conduct experiments with accuracy and repeatability, freeing human scientists for creative discovery.^ix

As the trajectory of robotics rapidly evolves, tech and AI companies are amongst those looking into the integration of AI and robotics. With the global robotics AI chipset market projected to reach US$866m by 2028, industry leaders such as Boston Dynamics, Google DeepMind and Tesla are spearheading innovations that are reshaping the boundaries of robotics.^x Google Deepmind for instance, has announced its investment in research around training humanoid robots using LLMs, allowing robots to navigate their surroundings and have better spatial awareness. Google DeepMind has also spearheaded the Open X-Embodiment Collaboration, an initiative that has involved partnering with 34 research labs to gather data from 22 different robots.^xi This collaboration aims to provide researchers with access to larger, more scalable, and diverse datasets. The Toyota Research Institute, in partnership with Columbia University and MIT, is exploring a technique called diffusion policy.^xii This approach aims to teach robots to quickly and easily acquire new skills and has already equipped them with over 200 dexterous skills—all without requiring any new code to be written!

Aside from larger companies, numerous startups are also looking to capitalise on the surging popularity in robotics. Venture capital firm Sequoia Capital has invested recently in a number of AI robotics start-ups, including RobCo and Collaborative Robotics.^xiii

With the potential for autonomous systems to orchestrate both digital and physical realms, we are on the cusp of an era where intelligent machines become economically viable and integral to everyday life. However, while the landscape of robotics is certainly exciting, it is important to consider limitations such as questions around accountability, safety and bias in decision making, and environmental concerns around solutions to data shortages. Through responsible and transparent development, addressing social concerns, and prioritising sustainability, the rise of intelligent machines holds the promise of increased efficiency, expanding market opportunities and unlocking o innovation in untapped industries.

Conclusion

GenAI technology is constantly developing and iterating. With ongoing advancements in self-improvement and reasoning, AI systems are becoming increasingly adaptable, reliable and impactful across diverse domains and industries.

Pivotally, the convergence of AI and robotics marks a transformative moment in technology. Integrating LLMs into robotics can significantly lower the cost of physical manipulation, making sophisticated robotic capabilities economically feasible and more accessible to enterprises and consumers. This new frontier invites us to imagine a world where AI-driven systems bridge the gap between digital and physical domains, crafting solutions we’ve yet to envision.

_____________________________________________________________________________