This article explains our thought about the usability of artificial intelligence in paper and forest industry. The author, Heikki Peltonen, is the CEO of Vipetec Oy, a company that has been working more than 20 years in data-driven paper and forest industry process development. The results have been significant and clearly measurable. Before launching Vipetec Heikki was a university researcher in this same field. This gave him an insight into the quality and quantity of the information that a single paper machine can provide. He also witnessed the decline in the hype of neural networks in process control at the turn of the century. Here Heikki shares his views about where we are compared to those days.
Currently there are several artificial intelligence projects in paper and forest industry. Is it possible to find a general solution through them to these processes? After all, computing power, algorithms and methods have evolved over the years. Challenges here still include the reliability/validity of the data and human errors. The old statement in modeling, “garbage in, garbage out”, seems valid even today.
What Heikki has learned during the 25 years about the challenges of using and refining data:
1. It is good for the user to have an understanding to what, for example, the guidance proposal calculated by data analytics or a model is based on and what limitations it may have.
2. The results of the calculation must be reliable with very few errors or suddenly their use is negligible or ceases.
3. The benefits must be measurable, a gut feeling is not enough.
These points alone place an extra burden on large artificial intelligence projects.