S7E153: AI in Project Controls: Separating Fact from Fiction with Alan Mosca

<p>In this week’s pod, we welcomed back Alan Mosca to discuss AI in Project Controls – Separating fact from fiction.</p> <p>Alan is the co-founder and CTO of nPlan, where he leads technology, research, and product, whilst developing thought leadership about forecasting and risk. Before nPlan, Alan spent 7 years as a technologist in quantitative finance, on live trading systems, research, and front-office in both high-frequency trading and asset management.</p> <p><br> Alan has extensive experience in algorithm design and software engineering and holds a BEng in Computer Engineering, MSc in Computer Science, and doctoral research in machine learning theory.</p> <p>The main topics we discussed on the podcast were as follows:</p> <ul> <li>There needs to be a lot of responsibility with using data from AI toolsets</li> <li>Toolsets are available that can auto-plan a successor activity in a schedule based on previous data</li> <li>Large language models only work based on language. Chat GPT took longer to reach the mainstream because of the data checks to ensure outputs were not inappropriate</li> <li>AI will not replace humans on projects. It will only evolve their current roles</li> <li>In the next 3-5 years we will see models that can internalize the meaning of a project</li> <li>AI could be used to measure schedule compliance with the contract</li> <li>Models can’t self-regulate which can lead to biases in data.</li> <li>We’re past the point of having a common data environment</li> <li>Create better things not faster things!</li> <li>Simulation is harder than AI because it requires a greater level of precision</li> <li>One person’s experience is another person’s bias</li> <li>One of the main fictions of AI is that everything will be possible. It will never predict the future, it will only forecast possible outcomes</li> <li>Be a critic! AI outputs are not infallible</li> </ul> <p>Here are links to some of the topics we discussed:</p> <ul> <li>David Chalmers – Are Large Language Models Senti