About AIARD

Inventing is a scientific problem, and we treat it as one on both sides. On one side, inventive design: the theory, derived from TRIZ, of how difficult technical problems are solved by identifying and overcoming contradictions. On the other, computer science: language models, information retrieval, text embeddings and benchmarks.

The theory keeps the AI honest. A language model gives an opinion; a contradiction, a problem graph or the Altshuller matrix gives a structure that can be checked. Our question: can a machine help invent, and can we prove it?

Inventive design has been developed at INSA Strasbourg since 1995, as an extension of TRIZ that could be reproduced, taught and computerised: contradictions, problem graphs, laws of evolution, ontologies for computer-aided innovation, published in CIRP Annals, Computers in Industry or Expert Systems with Applications. Work on combining TRIZ and artificial intelligence began in 2009.

Strasbourg hosted the CAI 2011 conference and the TRIZ Future Conference 2018, and since 2020 the team has co-edited the Springer proceedings of the field’s world conference, renamed in 2024 the World Conference of AI-Powered Innovation and Inventive Design.

The AIARD industrial Chair was launched on 26 November 2020 with five founding patrons, the support of the Région Grand Est and the Eurométropole de Strasbourg, and funds raised through the Fondation Université de Strasbourg.

Since 2021, six doctoral theses have been defended in the team and two are under way. They produced resources the AI community uses: patembed, twelve patent-embedding models downloaded more than 23,000 times; PatenTEB, a 15-task benchmark; DAPFAM, a cross-domain patent retrieval dataset now part of MTEB, the reference embedding benchmark.

Earlier theses brought TRIZ to computer-science venues — contradiction mining in patents at IEEE ICDM and ICMLA, solution ranking at DEXA — and work published in Scientific Reports links engineering problems to biological solutions across more than 101 million abstracts.

Each of our tools comes out of a thesis: PatentSpark, FusionMind, HiveScan, ARIZ-Copilot, Finder and TRIZBench. They go to industrial members first, in live demonstrations.

The Chair is funded mostly by industry. Industrial partners support it through annual patronage; in return they take part in its activities and give IDEAS accounts to their teams, which open all the Chair’s software. Patrons change over the years; ArcelorMittal and Forvia have been with us from the start.

Public partners support the Chair as a research capability for the territory. Expertise assignments are contracted separately through SATT Conectus.