GenAI: Is it all about Attention or also about Predictability?

Einleitung/Dauer

Artificial Intelligence is rapidly moving from experimentation to infrastructure in actuarial work. AI systems are beginning to influence decisions that were historically driven by statistical models, expert judgment, and regulatory constraints. This session focuses on understanding what is happening under the hood of modern Generative AI systems, particularly large language models and AI agents. What does “attention” mean in technical terms, and why is it foundational to how these systems process information? How do agentic systems differ from classical predictive models? And critically for actuarial practice: where does predictability break down?

We will examine both the capabilities and the limitations of AI. In domains characterized by uncertainty, feedback loops, and human behaviour, no system, human or machine, offers perfect foresight. Understanding these boundaries is essential for responsible adoption. The objective is not to replace actuarial judgement, but to augment it, while ensuring that humans remain accountable for decisions in high-stakes contexts.

Preliminary Programme

Friday, 9 October 2026
10:00-11:00 All about attention and why it matters much for AI agent training
11:00-12:00 Agentic applications and why predictability is a different matter

All the above times are given in CEST (Central European Summer Time).

Vorgehensweise und Ziele

We aim to deepen the understanding of how GenAI models work and why and how they understand digital context in the way humans understand broader context. We also want to shed light on the limitations of GenAI capabilities, as well as human limitations, when it comes to predictability of systems of complex and dynamic nature, which occur in nature, societies and finance and why humans therefore need to stay in the loop when AI work results are used for that matter.

Teilnehmer

• Actuaries (junior to senior levels) – To understand how their role is evolving and how they can actively shape AI initiatives
• Heads of Actuarial & Risk Departments / Chief Actuaries – To explore strategic opportunities for AI adoption and lead cross-functional initiatives
• (Actuarial) Data Scientists & AI Specialists in Insurance – To gain insights into use cases best suited for the qualities of large language models
  

Technical Requirements
Please check with your IT department if your firewall and computer settings support web session participation (the programme Zoom will be used for this online training). Please also make sure to join the web session with a stable internet connection.

Dozierende

Dr Noa Zamstein
Noa is a Senior Data Science Researcher at Earnix, where she develops advanced analytical and modeling solutions for the insurance and banking sectors. She holds a PhD in Chemical Physics and brings over a decade of experience as a data scientist. Noa is the recipient of the 2025 CAS Ratemaking Prize, awarded for her synthetic data solution for privacy preservation.

Marcus Looft
Marcus is a qualified actuary with 25+ years of experience in insurance. He held leadership positions with Allianz and AXA, and consulted for Earnix, Milliman, and WTW. He co-founded Simpego, an insurer in Switzerland that uses AI-driven market pricing and cloud-based operative systems. He's currently Director of Strategy for Earnix.

Sprache/Kurztitel

The language of the web session will be English.

CPD Credits
For this web session, the following CPD credits are available under the CPD scheme of the relevant national actuarial association:

• Austria: 2 points
• Belgium: 2 points
• Bulgaria: 3 points
• Croatia: individual accreditation
• Czechia: 2 hours
• Denmark: 2 credits
• Estonia: 2 hours
• Finland: 2 points
• France: 12 points
• Germany: 2 hours
• Greece: 3 points
• Hungary: 2 hours
• Iceland: 2 credits
• Ireland: 2 hours
• Italy: individual accreditation
• Latvia: 2 hours
• Lithuania: 2 hours
• Netherlands:  approx. 2 points (individual accreditation)
• Norway: 2 points
• Poland: 2 hours
• Portugal: 2 hours
• Serbia: 2 hours
• Slovakia: individual accreditation
• Slovenia: individual accreditation
• Spain: CAC: 2 hours, IAE: 2 hours
• Switzerland: individual accreditation
• USA: SOA (Section B): up to 2.4 hours

No responsibility is taken for the accuracy of this information.