Beyond Gen AI
Engineering leaders need more than Gen AI alone can offer. They are inherently focused on solving real-world problems. They want to harness the power of AI to transform processes and also to create game-changing smart new products and services. But it needs to be AI that is fit for purpose – AI that can meet their demands for accuracy, specificity, quality, compliance, and dependability in product, not just desktop environments. As we’ll see, Gen AI in isolation may help to achieve some early results, but it simply can’t always satisfy these rigorous criteria. Thinking that does not deal with the specific failure modes of generative technology, will likely miss key engineering requirements.
Engineering requires both precision and the ability to capture detailed data in many forms. There is a vast difference in engineering application between a photo-realistic video and the schematic diagrams of an airliner avionics system, even if both are represented graphically.
The requirement extends beyond text, image and video. Engineering organizations need:
To factor in engineering diagrams, architecture, and process flows into the corpus of inputs and outputs of AI models
To control precision, and the correctness of the system
To be able to gauge the confidence levels in what AI is telling them
To balance the innovation of AI against engineering’s heightened requirements for things like safety, regulation, intellectual property, privacy, and security
To act fast – but also, to build strategic foundations over the long term
To embed AI not just in processes, but in trailblazing new products and services
To achieve balance between the ability of AI to transform against sustainability commitments. For instance, ever larger language models require ever larger compute facilities to train them, which have significant energy requirements
This is why Capgemini Engineering’s approach combines generative machine learning models, which characterize patterns in data, with sources of structured knowledge and symbolic reasoning, including traditional symbolic Al, knowledge management, reasoning, and planning. We call it Hybrid AI.It’s an approach that means the value of AI is delivered with the quality and correctness the sector needs. Implemented on an industrial scale, it will change the way we approach mainstream engineering. It won’t replace human endeavor, but rather enhance and accelerate it. It will, in other words, become Augmented Engineering – taking people, processes, products and services to new levels of intelligence and efficacy, while retaining both rigor and creativity.
