Developing your AI project: easier than you think
No need for powerful computing machines – Contrary to popular belief, these days, machine learning algorithms can be trained, tested and deployed in just few minutes using big data and at a lower cost, thanks to cloud services and graphical processing unit (GPU) servers. This eliminates the need for hardware infrastructure (e.g., computing servers, virtual machines) to design AI-based solutions.
No need to be a programmer – Thanks to advanced data analysis platforms like Microsoft Azure ML and Amazon SageMaker, you can quickly and efficiently develop, test and deploy many AI models—classification, regression, clustering, recommendation—without even knowing how to code. Moreover, given the availability of so much public data on the web, also known as open data, you can test these algorithms and see how efficient they are without having to invest in data collection and structuring.
And if you want more flexibility – Free libraries like scikit-learn, TensorFlow, PyTorch and Keras, just to name a few, can help you to develop your own AI algorithm using existing features.
Data – Most current databases are not organized enough to be directly used in an AI project. It is generally critical to efficiently manage and handle data, which generally accounts for 80% of the time and effort spent on an AI project. It is important to ensure not only that the right amount of data is available, but also that it is relevant and of good quality.
AI model expectations – As a user of an AI-based model, you must ask yourself the following questions: How reliable is this model? Does it meet my expectations? What do we want to improve in our daily tasks?
Model control and maintenance – An AI-based model can solve existing problems. However, it needs to be monitored and maintained over time to preserve sustainability and resilience. For example, as new data accumulates, it is important to retrain the machine-learning algorithm using this data and to observe the results. Sometimes, you need to make some adjustments to keep a certain degree of accuracy in the algorithm.
First, you must bear in mind that AI is only a tool to help people work, solve problems and assist in decision-making; in no way can it replace the human expertise required to make your project successful. As a result, you cannot blindly rely on AI results.
Another important factor to ensure a successful AI project is understanding the data and having a good strategy to manage data well. The better you understand and manage your data, the more successful and reliable your AI project will be.
You also need to consider the sustainability of the information you gather by asking yourself if the data you collect today will still be usable tomorrow. It may be time to reorganize your database and realign your future data collection. For example, consider collecting quantitative rather than qualitative data. Also, keep in mind that in a few years, in-house software may no longer be used, so it may be difficult to retrieve the data from this software.
Finally, for your AI project to be successful, rely on a multidisciplinary team: AI algorithms are as complex to understand for a geologist/ mining engineer as the geological context or mining plan are for an AI specialist. For example, if you are carrying out a geology project, and you have no idea what “geological setting” means or what “artificial neural networks” mean, you do not have the right tools to carry out your project. The expertise from both these types of specialists is necessary to successfully apply AI to your projects.
Examples of AI applications in the mining sector
There are many AI applications in the mining sector. Here are some examples:
- Analysis of labour or machine productivity and automatic notification of detected issues;
- Autonomous vehicles and drilling machines to lower costs and increase employee safety;
- Analysis of geological information to find the best drilling targets using regression;
- Automatic mapping of an area based on data already collected using classification;
- Automatic description of drill cores or rock samples through a classification process;
- Automation of extraction operations to increase ore quality, throughput and recovery to maximize production through visualization and automatic classification of rock into ore and waste rock;
- Using cameras for remote control of your mine site to monitor critical equipment or processes using artificial intelligence-based video and image analysis methods
- Predicting failures or other potential hazards by analyzing event trends.
University research and many industrial studies on AI have led to significant improvements in algorithms and to introducing new, more sophisticated and robust AI models, like deep-learning algorithms. Take advantage of this knowledge to develop highly precise AI-based solutions.
There are also many sources of funding to support your AI-related projects and initiatives, including from governments or public and private sector companies. These subsidies make it easier to carry out and support AI projects.
Ready to put artificial intelligence to work for your business, but unsure about how to get started? Consider our AI Jumpstart Program. Over 3 to 4 weeks, the program will help you see where your organization stands in terms of artificial intelligence. You’ll also come away with a clear roadmap on how to achieve your goals and maximize your return on investment as quickly as possible. Contact us to find out more.