robot in factory
getty
Artificial intelligence aids manufacturing by better understanding product demand and input supply, while reducing the impact on what's happening inside the factory. Robotic technology continues to improve, but not at the same rapid pace as the large-scale language models (LLMs) that power chatbots such as ChatGPT.
LLM has made great strides in processing words, numbers, and images. Many people think of “robots” when they hear “artificial intelligence,” but don’t expect LLMs to power robots anytime soon. LLM showed its power when large amounts of data collected from the Internet were used to train models. Comparable data have not yet been collected regarding robots. Robotics will continue to improve thanks to better sensors, chips, and experience, but it won't be progressing at the fast pace of an LLM any time soon.
The benefits of AI in manufacturing will primarily come from a deeper understanding of product demand and material supply. Both benefit from integrating large-scale language and statistical models. Businesses have been using statistics for many years, but the data needed to be formatted accurately and data analysts needed to know the different types of predictive models and how to use each. Analysts then spent hours guessing at the right model to use or trying out different models to see which one worked best.
One model of predictive analytics systems combines large-scale language models with statistical analysis software. LLMs receive instructions in English such as “Please help me predict how many Widget-1000s will be ordered next July.” The system then connects LLM to the company's database of past orders. LLM can format the data as required. You can develop statistical models or apply machine learning directly to your data. Finally, the system can communicate the expected order quantity and 95% confidence interval to the factory manager in English or any language.
On the supply side, a similar process can predict late delivery of input materials and worker sickness and vacation.
The predictions aren't perfect, but they're better than the typical “fly in the wind” approach.
Engineers will find AI useful in product design. Research firm AIMultiple explains: “Designers and engineers input design parameters (materials, size, weight, strength, manufacturing methods, cost constraints, etc.) into generative design software, and the software provides all the results that can be created using those parameters. Parameters. This method allows manufacturers to quickly generate thousands of design options for a single product.”
On the factory floor, two AI advances are improving activities. Machine learning (a component of AI) can help predict required maintenance. In addition to the maintenance schedule recommended by the manufacturer of the machine, the system can monitor the performance of both quantity and quality of production. This data can be correlated with past downtime records to predict future issues. This is easier in factories where there are many machines of the same type.
Another factory floor improvement involves automating quality inspections. While some tests can be performed with simple, older technology, some of today's tests still require human judgment. Such products are inspected using sensors. Inspection can check image size, defects, weight, sound, and more. The system will be trained at the same time as humans are inspecting the process. The system tries to match human judgment. Eventually, systems will become just as good as humans. Human inspection could then be performed to verify whether the AI is working correctly.
AI will aid manufacturing, but it will not revolutionize it. This is consistent with my general conclusion that the realm of physical objects is less affected by AI.
Look out for further gains in the coming years.
follow me twitter Or LinkedIn. check out my website.
