Conference Presentation, Keynote, Product Demonstration
Keynote by Caspar Herzberg, AVEVA Systems CEO | RAISE Summit 2024 | Paris
- Leveraging AI to eliminate human error in industrial processes is projected to achieve energy consumption reductions and efficiency gains of 20, 30, 40, or potentially 50 percent.
- Industrial data volumes generated over the past 30, 40, and 50 years are expected to continue growing exponentially, though more than 70% of existing data, including time series and sensor information, is anticipated to remain unused.
- Currently, only 40% of global companies are expected to possess the capability to successfully access and utilize portions of this industrial data.
- Organizations that effectively share data across supply chain ecosystems are projected to outperform non-collaborative peers in growth, profitability, and overall performance.
- A persistent human reluctance to share data and automate, driven by psychological needs for protection within large organizations, may limit the success of general AI and automation.
- Challenges regarding inconsistent data, varying formats, and a lack of collaborative spaces are expected to be addressable through the implementation of proper infrastructure.
- Large language models are predicted to generate errors at a rate comparable to solutions, necessitating significant human insight to distinguish correct outputs from mistakes.
- Specific reliability of the AI application is expected to be near non-existent regarding failures due to limited input scope, such as machine time series and manuals, which reduces human error or machine hallucination risks.
- The company plans to launch a new AI product in June that integrates operational data, predictive analytics, and an industrial assistant.
- This new product is expected to reduce the training period required for engineers to support decision-making from a traditional 12 to 14 years down to one or two years.
- Aveva is expected to create a product combining large language models with predictive analytics, serving as an early real-world industrial AI application using a "modest approach" to address specific use cases with a focus on zero faults.