Salt, Pepper & AI
“HAL 9000”, the sentient computer imagined by Arthur C Clarke in his remarkable sci-fi classic “2001: A Space Odyssey”, locks the astronauts out of the spacecraft because it concludes their actions could jeopardize the mission, it is a stern reminder of the potential benefits and the dangers of AI.
There was caution and excitement about the arrival of generative AI, some showcasing how AI accomplishes certain tasks which are usually done by the beings of flesh and bone, at the same time there were calls to stop any further work until more ethical frameworks were put in place to govern the AI models.
I was fiddling with the code of a sample AI model in Colab, the dataset used to train the model contains parameters usually observed to asses whether a breast tumor is benign or malignant, it occurred to me that, maybe the job of a radiologist could be most supplemented by AI in medicine. A well-trained AI model can make instant and accurate predictions, better than a radiologist with several years of experience behind. This could be a game-changer.
Think of the salt and pepper shakers we find on restaurant tables; standalone they have no utility; however, if you supplement your dish with them in the right amount, they greatly enhance the flavor and taste. At this point, our lives are going to be greatly enhanced by supplementing AI to the right levels on the stuff that matters most to humans.
There is a lot of talk about how AI could enhance productivity in many domains. In my view, the below aspects of the manufacturing industry can be greatly enhanced with the adaptation of generative AI.
Design
Despite the mass adaptation of computers & software that aided the design process in the last couple of decades, this is still highly iterative and follows the “design, analyze, build, test, repeat” cycle, in a linear fashion at that. A successful design depends a lot on human intelligence and domain experience. There were some advancements in the last few years with topology optimization software, that helped fine-tune the design for components, however, topology optimization failed to take off en masse owing to the limitations, that these designs, by and large, cannot be manufactured using conventional manufacturing methods. Here, generative AI is giving some hope, with companies like PTC and Autodesk rolling out design tools that help engineers find the optimum design, reducing iterations, which also can leverage conventional manufacturing to produce them. The silver lining is the ability of these tools to iterate the designs in a non-linear fashion with the help of algorithms powered by neural networks. This could greatly speed up the design process in the coming years, providing much-needed respite for engineers and companies alike.
Project Management
The time it takes to deliver a project primarily depends on two factors, the organization’s capability, and its supplier’s capability. Most of the time the schedule and cost overruns that hurt the most arise from the misunderstanding of these two at the start of the projects. The ideal situation is, the project manager factors these two comprehensively while charting the project. However, this is easier said than done, as it depends on the data from past projects, such as resource constraints, and insights from missed deadlines, and so on. Even if these data are captured and made available, it is near impossible for a human brain to crunch all these and use them to develop good project plans. This is something that can be easily handled by an AI model that was trained in those datasets. they can provide insights for accurate task durations, spot risks, and suggest concurrent paths. Besides, project managers spend a lot of time updating dashboards, writing reports, and organizing meetings. Most of these can be offloaded to an AI assistant. This could roughly cut 50% of the time spent by project managers which can be used to communicate with project sponsors, resolve risks, and coach the team.
Commodities
You may now catch the falling knife, will you? Commodity price predictions have always been challenging. Qualitative analysis can provide more accurate predictions however they are limited to only near-team predictions and subject to human bias. The quantitative predictions using AI models could make a lot of difference for organizations that want to mitigate supply-side risks of raw materials. For example, a computer can process a large amount of data and the algorithm can factor in macro and micro-economic data, and recognize patterns. These are near impossible for a human brain to handle. On top of this, these ML models can get a lot better with time as these models can be continuously trained, especially when they make a wrong prediction, this improves the accuracy of future predictions. This will be a boon for organizations to manage supplier risks, cost modeling, spend forecasting, etc.. much more accurately, reducing human dependence and associated errors.
In 1930, John Maynard Keynes predicted his grandkid’s generation will be working 15 hours a week, given the advancement in technology. That prediction never came true, production and consumption have increased significantly since then. Not something Keynes, who is regarded as the founder of modern macroeconomics anticipated. Now, we are entering the next level of advancement with AI, impacting life as we know it today. Using this to solve some of the complex problems the world faces and promote sustainable living should be of priority, this responsibility summarily falls on the shoulders of this industry.
