Recently, Vinod Khosla, the founder of Khosla Ventures, and an initial investor in OpenAI, emphasized why robotics technology will soon experience its AI breakthrough.

Khosla states that the transformative capabilities of AI foreshadow a future where AI and robotics technology will liberate humans from mundane tasks.

He says that robots will experience a "GPT moment" within the next 2-5 years, where robots will transition from being programmed (following instructions) to learning systems that understand the dynamics of the physical and real world, thus advancing robotics technology rapidly.

Already Happening

A few days ago, NVIDIA researchers introduced DrEureka, an agent powered by a large language model (LLM) that can automate processes simulating real-world scenarios, effortlessly training a robotic dog to balance on a yoga ball without fine-tuning.

Interestingly, DrEureka is built upon their previous work, Eureka, an algorithm that teaches a five-fingered robotic hand to spin a pen. "This is another step in our pursuit of automating the entire robotic learning process using AI agent systems," said Jim Fan, Senior Research Manager and Head of Embodied AI (GEAR Lab).

Figure 01, powered by OpenAI, has also made significant progress in visual reasoning capabilities. Recently, it has been able to differentiate between healthy options like oranges and less ideal choices like potato chips, mapping camera inputs to robot actions at a rapid rate of 10 times per second using its internally trained neural network.

Brett Adcock, the founder of FigureAI robots, believes that "everyone will have a robot in the future, just like owning a car or a phone today," he added.

Tesla is not falling behind either. Recently, Optimus is ready to work in factories, utilizing the Full Self-Driving (FSD) computer to classify battery cells in real-time. It can precisely classify battery cells with minimal insertion tolerance and automatically locate the next available slot.

This year, Google DeepMind also released three robotic research systems - AutoRT, SARA-RT, and RT-Trajectory, which will help robots make decisions faster, better understand and navigate their environments. These models will aid in data collection, speed, and generalization.

In addition, Stanford University has introduced the Mobile ALOHA system, which aims to replicate dual-hand manipulation tasks that require full-body control.

Supported by Google DeepMind, this technology addresses the limitations of traditional imitation learning from human demonstrations. These versatile robots have proven to assist in various tasks such as cooking, cleaning, weightlifting, and other manual activities.

What's Next?

While advancements in artificial intelligence research are still commonplace, major companies are racing to seek the next breakthrough in robotics technology. Just like NVIDIA's GR00T project released a month ago, followed by Dr Eureka, as mentioned earlier, more companies are investing heavily in robotics technology.

Major players like Google DeepMind, Tesla, and NVIDIA prioritize robotics technology, so significant breakthroughs may occur soon. Significant progress has also been made in open-source research, with Hugging Face launching LeRobot, an open-source robot database a few days ago.

As correctly pointed out by NVIDIA CEO Jensen Huang, "For leading roboticists worldwide, various empowering technologies are converging to drive tremendous leaps in AI general robotics technology."

Clearly, the ChatGPT moment for robotics technology is not in the future but already here!