The fascinating (new) world of artificial intelligence (AI) and thereby especially machine learning (ML) is for Knut Höngesberg already over a long period of time a very interesting one. This applies in particular to exciting applications within corporate environments.

He is thereby less interested in populist discussions about 'autonomous robotic beings that take over humanity'. Instead, it is the extraordinary potential of supportive and automating AI/ML that draws his attention most, i.e. those applications that simplify people's lives, that massively increase the quality of life for people, as well as the opportunities for increasing efficiency in companies. In his opinion, the latter have the strength that corresponds in historical comparison to huge changes like the introduction of mechanical looms or industrialisation.

Academic education:

In 2019 and early 2020, Knut also extended and directed his lifelong learning about the AI/ML subject on a more academic footing. For that reason he spent several months studying i.a. the course material "Machine Learning" authorized by Stanford University, California/USA. He also took part in all the associated exams and was able to pass them with a rating of 98% (course certificate "Machine Learning", offered through Coursera.org and authorized by Stanford University)

Main focus:

Knut focuses on the topics of supporting AI and automated AI (i.e. not autonomous AI). He provides answers for typical related challenges of companies in these times - on the one hand to deal intelligently and creatively with the 'flood of data', but on the other hand also knowing that there can be no intelligence without data.

Therefore, he especially uses machine learning (ML) including deep learning (DL) and neural networks (NN). Of course, in some cases there is an overlap with the topics of data science (DS), where results are more created by algorithms rather than from data like with ML.

In machine learning, he covers both the topics of supervised learning and unsupervised learning.

Currently, he is thereby using actively the i.a. following methods and models:

• Classification and regression models
• Linear regressions, i.a. with many hundred of variables (multivariate linear regression)
• Logistic regressions (e.g. with Sigmoid), thereby determining linear as well as non-linear complex boundaries
• Neural networks, modeled on the approaches of the human brain (including feedforward propagation and backpropagation)
• Support vector machines (SVM)
• Unsupervised learning with i.a. K-means algorithms and clustering (e.g. for downstream reasons from a business perspective)
• Status determination and dealing with overfitting and underfitting of mathematical models, as well as regularisation
• Methods of dimensionality reduction, e.g. for data visualisation
• Anomaly detection
• Collaborative filtering (recommender systems)
• AI/ML system diagnosis and sanity check, treatment of the 'cocktail-party problem', and the like
• Machine learning pipelines and ceiling analysis
• and from a performance as well as computational expense perspective, of course, the art of linear algebra with vectorisation and matrices, stochastic gradient descent, mini-batches, parallelisation, etc

Summary:

As with his other IT-related topics, Knut Höngesberg aims on bridging the gap between business and IT/technology/software. For the sphere of artificial intelligence this covers strategic AI application aspects, associated project management or the requirements of a digital transformation board.

Likewise, a comprehensive understanding down to the (technical) level of detail is important and crucial for him. In learning and building AI/ML knowledge, he therefore also copes with all levels of design, prototyping and implementation (i.e. implementation roles like a machine learning engineer, AI product manager, data scientist, data engineer, etc) He is therefore able to create ML coding and prototyping examplarily by his own (using e.g. a prototyping tool like Octave)

Further hints for reading:

At the end of 2019, I was invited to participate the congress of the eco Association of the Internet Industry. Under the umbrella motto "FAST.FORWARD.FUTURE" there were presentations and debates about the topic "Shaping the internet of the future with artificial intelligence". If you are interested in, find some of my notes and key findings out of this event over here, pls.

The best way to predict the future is to invent it.
(Alan Kay)