Five Robotics Failures and What Developers Can Learn From Them

OpenAI shifts away from robotics research

​OpenAI, which had researched and developed systems that could solve a Rubik’s Cube, decided to focus on areas of machine learning where data is more readily available than robotics.

The generation, collection, and annotation of massive data sets for training AI is a challenge for everything from autonomous vehicles to robotics manipulation.

Wojchiech Zaremba, co-founder of OpenAI, noted that the company has found its focus over time, now working on projects such as GPT-3 natural language processing. He noted that while robots can easily walk in simulation, doing so in reality is still more challenging.

Humanoid robots, fruit-picking robots, and the ongoing self-driving car race show that rather than blame luck or overestimate technology, designers can learn from past mistakes.

Return to article

Editors' Picks