AI creates value in the real world – also in industry

Raute Group

The coming of AI has been a recurrent topic in industrial discussions over the last years. Still, the discussion often revolves around expectations on artificial general intelligence (AGI), superintelligence, humanoids and alike.

To better understand the current status, we had a chat with Peter Sarlin, CEO, and Niko Vuokko, CTO at Silo AI, one of the largest private AI labs in Europe.

Q: Let’s start with a definition. How would you explain AI in layman’s terms?

Sarlin: AI is oftentimes defined at 3 levels: Superintelligence, AGI and weak AI. Today, there simply exists no technology for AGI or superintelligence which refer to artificial beings whose intelligence equals or far surpasses that of human beings. What we can see being applied successfully today is weak AI for narrowly defined specific problems. It relies mostly on a type of machine learning (ML) called supervised learning, the task of teaching a specific task to a machine with large volumes of labeled data. This is nothing more than a mapping function between ‘A’ and ‘B’, inputs and outputs. This allows identifying traffic signs or nearby vehicles in video images, giving music, book or video recommendations to specific users, translating sentences into different languages, combining spoken language with actions, among many others. And all of these technologies are examples of real-world consumer products, such as cars, web and mobile applications and phones. In short, AI today is creating value by automating or partially automating tasks as a part of everyday digital products.

Q: What does AI mean in the manufacturing context?

Vuokko: Supervised learning is the most common technology behind AI, and the main value creator in the manufacturing industry as well. It implies that we humans choose the relevant data sets, such as camera sensor, process, and quality data, apply them to a technical application, and approve the end result. When this process has been repeated sufficiently, the algorithm learns to repeat it. The change here with recent AI algorithms is the ability to handle more complex data and pursue more complex objectives.

As a complement to supervised learning, we see more and more applications also rely on unsupervised and reinforcement learning. This means that machines can also partially learn in an unsupervised fashion without examples and by reinforcement from automated rewards and penalties for performance on a given task. Example application areas for this are autonomous wood-harvesting and mining machines as well as automated manufacturing and production processes.

Q: When are we going to see the big transition to an AI-enabled world?

Sarlin: In the past 50 years, we've experienced multiple AI waves, mostly followed by AI winters. One of the first waves was in the 1960s and 1970s, after which they've been recurring every ten years or so. However, the current AI wave that started 5-10 years back has been different. Today, AI creates value as it is in everyday use and is present in day-to-day use – as personalized product recommendations, sentence translation, speech recognition, assisted driving, robot vacuums and lawn mowers, among many others.

So, AI does indeed impact business processes, and change organizations, the way we work and society at large. And this significantly impacts the way value is created in organizations. In particular, it is transforming how value is being created in digital products. But the next leap is to move from digital and technology companies to transforming more traditional industries. For these companies, the transition to an AI-enabled world will also require building a data infrastructure and a digital product that enables value creation with AI.

Q: What is going to change next?

Vuokko: AI thrives in making sense out of complex data. In manufacturing, as in many other fields, digitalizing different parts of the value chain or process phases has for now been done in separation. As this initial digitalization progresses, there’s now increased focus towards using AI for optimization across the whole factory and value chain. This ramp-up in the scale of AI investments is driving rapidly increasing focus on how to make these investments more efficient. Machine Learning Operations (MLOps), the methodology and tooling for efficiently scaling reliable operational use of machine learning, is the key enabler in getting a return on investments in AI. Another key driver behind MLOps is the spread of AI on the edge, that is, AI running within the factory as part of operations or safety critical functions. This shift has highlighted the need for a totally new type of expertise that combines in-depth skills in machine learning with everything from hardware knowledge to operational insight.

Q: Could you elaborate on the business impact a bit more?

Vuokko: Just to pick a few use cases, AI is already improving customer service, product design and performance, process optimization, asset performance, and helping to digitalize entire factories. Common to the strongest use cases is not using AI to provide insight or analytics for human consumption, but rather using AI to help machines make more intelligent decisions. While this potential is clear and tangible, it still often remains the future rather than the present simply due to the enormity of the vision. In general, the German automotive sector is a good example of taking determined steps towards the future. In short, they are not investing for short-term 3% gains, but rather for long-term 30% gains. They have also been mindful about the significant future progress in AI capabilities, thus avoiding lock-in with slow-to-move generic products. The lesson is clear for the Nordic industry: seek or build large value pools that can support the large AI investments that are going to be essential in this century.

Q: Where to start then?

Sarlin: Although it will take more than 10 years for ML to be a mature technology, it's now ready to be applied. What requires much more attention is identifying and defining a target use case that can be solved and creates significant value. And once that target has been identified, it's worth noting that the application of AI is an iterative process, where you learn along the way. It is a matter of true willingness and commitment; the value will follow. Most importantly, AI development should be approached as a long-term product development initiative that requires a proper digital infrastructure and persistent efforts and focus, rather than an ever-expanding list of AI experiments and trials.

Wherever you will start and however you will progress, it's worth noting that the future of AI will still rely heavily on the co-existence and co-operation of the machine and the human. But if that is done correctly, there’s a significant value creation opportunity for traditional industries like manufacturing.


About Silo AI – Silo AI is one of Europe’s largest private AI labs – a trusted AI partner that brings competitive advantage to product R&D. We build AI-driven solutions and products to enable smart devices, autonomous vehicles, industry 4.0, and smart cities. Silo AI provides its customers unique access to world-class AI expertise, as well as the Silo OS infrastructure to speed up AI development and deployment. Established in 2017, Silo AI is on a mission to build a European flagship AI company, with offices currently in Helsinki, Turku, Tampere, Oulu, Stockholm, Copenhagen, London and Palo Alto.