News Release

Mapping the cosmos of innovation: AI model charts the age and trajectory of 23,000 technologies

A new open-source dataset pinpoints exactly when innovations transition from the lab to the real world

Peer-Reviewed Publication

University of Technology Sydney

Radial Tree Dendorgam of Cosmos 1.0 and ET100

image: 

Radial Tree Dendorgam of Cosmos 1.0 and ET100. (a) The full radial tree dendrogram with two grey dotted circles demonstrates three and seven are optimal numbers of clusters for the TA23k. We use colours to distinguish the three meta tech-clusters (TC3). (b) The curated radial tree dendrogram shows the technologies of ET100 and cluster names of TC7 and TC3 from bottom to top. The circle sizes of ET100 represent the normalised value of a technology index called Generality_Index. The theme tech-cluster “Data & Analytics” is frequently mentioned across Wikipedia articles than other theme tech-clusters.

view more 

Credit: Gong, Xian, Paul X. McCarthy, Colin Griffith, Claire McFarland, and Marian-Andrei Rizoiu.

A team of researchers has built one of the most detailed open maps of emerging technologies yet assembled, allowing governments, companies and investors in the United States and worldwide to see what sits inside big fields like artificial intelligence and quantum computing, how fast each technology is growing and how deeply it is rooted in science.

Developed by the University of Technology Sydney (UTS) and partners, the Cosmos 1.0 framework uses machine learning to scan millions of Wikipedia pages, books and patents, then groups 23,000 plus technologies into a multi-level map.

Cosmos 1.0 also introduces four main indices that can be applied to any of these technologies:

  • Age of Tech estimates when a technology becomes part of everyday public life
  • Awareness tracks attention and public visibility.
  • Generality showing whether a technology is narrowly focused or used across many fields
  • Deeptech Intensity measures how strongly a technology is grounded in scientific research

Seeing inside AI, quantum and clean energy

The map lets users start with a broad label such as artificial intelligence or quantum technologies, then drill down into the underlying building blocks.

“For AI you can see the mix of technologies beneath the headline term, such as deep learning, transfer learning, computer vision and reinforcement learning, and compare how mature and widespread each one is,” said lead author and UTS industrial PhD candidate Xian (Elaine) Gong. “The same is true for quantum, where you can explore areas like post quantum cryptography, quantum sensing and different qubit designs.”

The authors found that the technology universe naturally groups into seven major clusters, namely: autonomous systems, biotechnology, data & analytics, health & medical, nanotechnology, networking & connectivity and converging technologies.

“We see six large clusters of technologies orbiting a seventh, central cluster of convergent technologies,” said co-author and UTS PhD candidate Claire McFarland. “It is where materials science, engineering and digital systems come together to create new combinations and hybrid innovative problem-solving technologies. Interestingly, many renewable energy and climate technologies sit within this core.” 

Deeptech, hype and strategic choices

The Deeptech index was designed to separate technologies that are deeply rooted in science from those that are mainly driven by marketing or business fashion. “The Deeptech index highlights technologies that draw heavily on scientific knowledge and specialised skills,” said co-author Colin Griffith. “Those are the capabilities that are harder to copy, take longer to build and often underpin national advantage in areas like advanced manufacturing, health and clean energy.”

By combining Deeptech with the other indices, decision makers can see which technologies are gaining attention, which are truly general purpose and which may be short lived buzzwords.

“With Cosmos you can generate rigorous ‘hottest new technology’ lists for any sector and see how mature each technology really is,” said co-author Paul X. McCarthy. “Because the dataset is validated against patents, venture capital investment and scientific literature, it can also support quadrant style charts that help leaders compare both individual technologies and groups of technologies by their sophistication and likely real world impact.”

The team, led by Associate Professor Marian-Andrei Rizoiu who heads the UTS Behavioral Data Science lab, demonstrated the capabilities of Cosmos 1.0. The framework can reconstruct the rollout of technologies over time in sectors such as automotive and mining, augment datasets to benchmark national strengths, and identify adjacent technologies that are natural next steps for existing capabilities.

The research is published in the journal Nature Scientific Data. The full Cosmos 1.0 dataset is openly available for researchers, analysts and policymakers to use in their own models, dashboards and strategic tools.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.