Artificial Intelligence is fundamentally altering human society. As AI takes over repetitive and dangerous tasks, humans are freed up to focus on more personally rewarding activities like those which require creativity and empathy.
This field has understandably attracted huge investments in technology development, with large players like Microsoft and Google focusing on this rapidly developing domain.
Where do we fit in? We believe we have a compelling alternative to the established orthodoxy of how ML systems are created. In essence, our approach replaces the 'obvious' neural network heavy approach with one which more closely models higher level human cognitive processes.
The foundational idea of heukno is a self-learning technology with applications to machine programming. The heukno technology distinguishes itself in two important ways from standard ML/MP techniques:
• It does not rely on existing datasets, but rather on self-generated data.
• The self-generated data allows explainability of the output.
Explainability of the output is highly desirable in many contexts, but elusive to standard ML/MP methods. In colloquial terms, the heukno technology can give you not only the answer, but also the reason for the answer.
The heukno technology has the potential to automate software generation in a number of areas which cannot be handled with current paradigms, e.g., maintenance of old systems with no available code databases for training, upgrading systems to new versions of the language in which they are written, and constructing/rewriting systems to meet given design guidelines. In short, the heukno approach to code generation is constructive, explainable, and adaptive, where current approaches rely on identifying and stitching together existing code snippets.
|
|
Contact:
heukno
DTU Science Park
Diplomvej 371
2800 Kongens Lyngby
Denmark
Google Maps