AI100 report released: embodied AI highlighted as grand challenge

The One Hundred Year Study on Artificial Intelligence (AI100) is a longitudinal study that assesses AI’s development and societal impact. Reports are released once every five years, with the 2026 edition the third in the series. The title of this year’s report is “AI Goes Mainstream: The Scaling Era and the Rise of Generative AI”. The panel behind the project is an international committee of experts tasked with examining AI’s evolution across technical, ethical, economic, and societal domains. They respond to 12 standing questions about AI’s advances, opportunities, dangers, and societal effects, and compare developments since 2021 with the previous panel’s assessments.

The 2026 AI100 Study Panel report examines the scaling of data and computing resources behind generative AI, advances in scientific discovery, investment in AI infrastructure, the concentration of frontier AI development, cybersecurity and military applications, the trustworthiness of digital media, emerging risks, public participation, and international governance.

One of the standing questions concerns open grand challenges in the field, with the experts highlighting six as particularly inspiring and scientifically important. Most relevant to the robotics community is the problem of physical AI and embodied intelligence. In the five years since the last AI100 report (published in 2021), we’ve witnessed integration of natural language processing (NLP) into robotic systems, the use of vision-language-action models for training and control, and advances in simulation technology which have allowed models trained in virtual settings to be deployed on physical hardware.

The 2026 AI100 report notes that China is the leader in industrial robot development, deploying more robots than the rest of the world combined. China also produces about 80% of the world’s humanoid robots. Although this sector currently accounts for just 2% of the market, it is growing rapidly.

Despite the progress of recent years, there remains a huge gap between the demonstration of capabilities and deployment. Many of the success stories have been in highly constrained environments (such as warehouses), or with very specific products (for example, floor cleaners). The key challenge remains the sheer complexity of the real world. Some of the scientific problems to overcome include long-horizon planning under uncertainty, achieving dexterous manipulation of diverse objects, and establishing formal performance guarantees.

Whilst there are some practitioners who believe that the current statistical learning paradigms are sufficient for attacking these open problems, many others feel that fundamentally new ideas will be required to bring robot deployment into genuine real-world scenarios. It will be interesting to observe how this plays out over the next five years.

You can read the report in full here.