Artificial Intelligence in Manufacturing – The AI World
AI has many strengths that we can use in our manufacturing and our R&D operations. But AI is not the same as a thinking human brain. It is a tool for the humans, but it does not replace them.
Before Christmas 2024, Ian Bremmer, founder of the Eurasia Group, predicted that within 3-5 years the most important relationship we’ll have will be with our AI (probably in our phones), and this personal AI will ‘know’ us better than anyone else. And by anyone else, he meant family, spouse and employer.
Bremmer’s forecast means that in 10 years the role of government becomes not only disintermediated – as individuals and businesses increasingly use their AI to interact with each other – but paradoxically the citizen-government relationship also becomes more centralised as the control and regulation of IT and data occurs in ever-fewer places.
In this paradigm, the traditional role of ‘government’ is not needed so much for regulation of interaction between people – AI does it. And this opens an opportunity for bad actors, especially foreign governments who – for instance – want to influence voters in an election. Canadian intelligence is warning citizens about this right now, as they move towards elections.
But whatever AI portends – positive or negative – how does its rise affect Australian industry?
Change agent – the good news is that Artificial Intelligence systems are already being used in Australian industry. They are deployed to drive efficiencies, aid design, achieve decarbonisation and control energy use. AI is embedded in existing ERP systems where machine learning, predictive analytics and natural language processing (NLP) are used. Australian industry uses AI in supply chains, in particular the automating of complex ordering/receiving and freight logistics, and management of complex energy systems: for example, where a factory has solar, battery and gas turbine and the owner is managing the optimum power output for the least carbon emitted. AI can speed back-office functions such as finance and HR and because it can be trained and customised, it can drive efficiencies in manufacturing processes.
Harrison Manufacturing and SPARC – at Harrison Manufacturing and our R&D company, Harrison SPARC (Special Projects and Advance Research Centre), we use a Microsoft pretrained Large Language Model based on SharePoint and OneDrive data. We are able to source enormous amounts of data, but we do it by separating the systems into two distinct entities: on-premises, and cloud-based (AI) language models. Our on-premises model handles sensitive internal operations within our secure environment, whereas we leverage cloud services like Microsoft’s Copolit for general-purpose, non-sensitive tasks. These systems operate independently for security reasons and to minimise sensitive information leaking into the public realm. We use retrieval-augmented generation (RAG) to enhance our models’ responses, by grounding them in our internal documentation and knowledge base, which we are constantly growing. By using RAG our on-premises AI can collaborate with our researchers, giving them the up-to-date information on the latest trials and experiments being performed.
The future of AI – the Economist says 75% of AI use is currently personal, ChatGPT etc. The rest is corporate which is the category we fall into. Clearly while consumers blaze the trail, it will be businesses and governments that put AI to the most powerful uses in the future. Ian Bremmer says AI will fill the intermediary roles currently conducted by governments, giving rise to the “Gzero” scenario where – rather than the G20 or the G7, or COP – governments won’t need to meet and align on policy directions. Global relations will become more regional and fractured, and assisted by AI the countries will group themselves by national interest and perhaps on an issue-by-issue basis. Is this a useful trend? We have to suppose that it will be old-fashioned human judgement that decides that question.
AI downside – AI is not neutral. The people who build, control and train the AI program can slant it in the direction they want. What could go wrong? Well, you could rely on an AI app to show you an image from history, but if you used Google’s Gemini AI in early 2024, you’d be struck by the fact that the AI would not return a picture of a European person. This concentration of data at one vendor might be powerful for a business owner or a manager, who is looking for operational improvements, innovations and new IP: such industrial breakthroughs are made with intense focus and being able to mine all of the data. But perhaps it isn’t so good for a consumer society that relies on IT platforms to search ‘facts’ and is instead served non-facts; if consumers are being asked to rely on AI, then it must be reliable. Consumers also need protection because most of them use the AI not knowing that the machine is ‘learning’ the user and pushing them into behaviours and purchases that they are not consciously aware of. The world’s most innovative people – inventors, engineers, scientists, writers, academics, musicians, directors and artists – are quite vulnerable.
They have a lot more to offer the AI vendors than normal consumers and many of these people have already had their IP and creative ideas taken by AI training models without their consent. We have to be careful here: there’s a big difference between Harrison Manufacturing collating all of its internal data and using AI to mine it… and stealing someone’s intellectual property to improve an AI program. And finally, while AI significantly improves our document processing and decision-making efficiency, it is important to acknowledge that by increasing our usage of AI over time we add to the well-known energy-consuming and heat-producing aspects of Artificial Intelligence. This puts us in an ethical dilemma: AI has the ability to summarise long documents so we can extract the salient points, speed our processes and decision making; but at what cost to the energy grid and the physical environment?
Artificial Intelligence is a tool for the humans – it doesn’t replace them.
AI is great for data retrieval, concept-synthesis and sorting through large sets of information to get answers. We use these AI strengths in our manufacturing and our R&D operations. But AI is not the same as a thinking human brain – our critical thinking, our reasoning, is still the most powerful thing. We see it like this: AI can improve human productivity gains when working in tandem with a team of great employees. That’s the quality we prioritise at Harrison – ‘Real Intelligence’ (RI). Artificial Intelligence is a tool for the humans – it doesn’t replace them.
