A summary of SIG History Event. "Simulation: From Humble Origins to AI Horizons," by Quintin van Heerden and Marno du Plessis

By Dave Evans

On 8 July at 18:00 Quintin van Heerden and Marno du Plessis presneted a SIG History event titled "Simulation: From Humble Origins To AI Horizons."

Quintin opened the talk by asking us what we thought a simulation was; amongst his possibilities was the idea that we live in one – a concept first brought up by the philosopher Nic Bostrom at Oxford University. Quintin gave a high level definition of it as being a modelling approach giving a representation of something which allows insights, scenarios, etc., to be explored.

Marno provided two pre-computer examples; one from 1777 being the Buffon’s needle example which generates a value for π (pi). A manual Monte Carlo model from the 1930s which generates pi was also covered. 

The general view is that significant simulation only began around World War Two.

The first major (discrete event) computer simulation language was IBM’s GPSS, in the early 1960s, with Marno covered with a model example. A high level summary was then given of developments over the following decades, covering the various different simulation paradigms which now exist, with examples of the various software packages available for each type.

Case study examples were described in areas such as logistics, aerodynamics, city evolution, population dynamics and healthcare. A summary was also provided of the seventy plus South African simulation case studies which had been either presented at ORSSA conferences or published in the ORSSA journal, Orion.

The talk then went on the address how AI is now being utilized in simulations: sometimes to enhance the model development process, and sometimes in agent system models, for example, where AI makes decisions. Several examples were covered. Whilst this is enabling major sophistication, it has significant risks too, with autonomous agents, etc., in the model which may not be doing what you think. This aspect, with validation and verification, came up substantially during the discussion – it was agreed that this had always been a slippery slope in simulation, long before AI, and could now be a bigger issue. A key message has always been that simulation must be decision support – not decision replacement…

In all, it was a wonderful overview of simulation. For more insightful talks like these, be sure to watch our event page (here).