AI for Materials Discovery | Konstantin Novoselov | ADIA Lab Symposium 2026

Value of the Information & Strength of the Argument

The talk’s value lies in its insider perspective from a leading researcher who has personally witnessed the full material-development cycle, from fundamental discovery through hype and denial to real industrial deployment. The argumentation is structured and empirical: each domain (transport, energy, infrastructure, electronics) is anchored in concrete quantified examples — 10-15% conductivity gains in super-copper, 1 kg/year graphene needed for all global touch panels, 20-50% concrete strength improvements. The reasoning builds logically from material production challenges to the epistemological limits of current AI databases, and culminates in a well-argued proposal to invert the predictive paradigm from structure-property to synthesis-property mapping. The boron nitride example provides compelling experimental evidence for the complexity this inversion entails. Throughout, the speaker maintains intellectual honesty about what remains unsolved, explicitly stating ‘we haven’t done any discovery yet,’ which strengthens credibility.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is high, reflecting the speaker’s Nobel-level expertise and direct laboratory involvement. The discussion of AI limitations is unusually candid for a keynote, acknowledging that existing databases are inadequate for real-world materials (steels, alloys, defects) and that synthesis remains a fundamental bottleneck. However, few explicit citations to specific papers are provided during the talk, and quantitative claims — while plausible and aligned with known literature — are presented without named references or data tables. The title adequately matches the content, which centers on AI for materials discovery with graphene as the illustrative thread. No comments were available for analysis.

The title accurately reflects the content: the lecture focuses on AI-driven materials discovery, using graphene and 2D materials as primary case studies.

8/10

Nobel laureate in Physics (2010) with direct research experience, presenting a balanced, honest perspective on both successes and limitations of AI for materials discovery. Claims are grounded in laboratory experience, though some quantitative assertions (e.g., concrete strength gains) lack explicit citation to specific studies during the talk.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 02:00Graphene introduction: structure, record properties, and the hype cycle from gimmick applications (e.g., $1M McLaren watch) to serious industrial use.
• 04:30Super copper with graphene additives: 10-15% conductivity increase enabling 4-5% efficiency gains in Chinese maglev train electric motors.
• 06:30Optronics as the promising future: optical interconnects, optical compute, and quantum optics; three graphene-silicon fabs under construction in Europe.
• 08:30Production methods: CVD and plasma technology; proposal to capture hydrocarbon flaring gas from petrochemical plants for graphene and hydrogen production.
• 11:30Remarkable scale fact: replacing all touch panels worldwide requires only 1 kg of graphene per year, motivating bulk applications in concrete and asphalt.
• 13:30Graphene composites: 0.5% graphene in concrete boosts compressive strength 20-50%, asphalt lifespan doubles, and concrete batteries for energy storage.
• 26:00Core AI challenges: databases contain only ideal equilibrium materials, missing real-world complexity of steels, alloys, defects and nanostructures.
• 33:00Boron nitride synthesis case: 33-dimensional reaction coordinate space, mapping intermediates and reaction pathways to enable targeted synthesis control.
• 36:00Knowledge-tree approach to hypothesis generation for high-entropy alloys and spin glasses; honest admission that no discovery has yet been achieved.

• ADIA Lab Symposium 2026 — Official symposium page mentioning the event schedule, speakers, and the session in which this talk was presented (Expert Case Study - Day 1: Prediction).

• Graphene — Wikipedia — Corroborates the described properties (strength, conductivity, thermal management) and applications of graphene mentioned in the talk.
• Chemical vapor deposition — Wikipedia — Confirms the CVD method described by the speaker for large-scale graphene production.