What are the uses of AI

Applications with Artificial Intelligence

Bernhard Zindel reports from the CAS AI Management class with Steffen Konrath:

After we had already had a few lessons about the functionality and mathematics behind Artificial Intelligence, as well as about the implementation and management of such projects, things are now getting hands-on.

The fact that the founder of AI Suisse knows where artificial intelligence is inside was clearly shown in his lessons. In addition to a huge collection of examples from all categories of artificial intelligence, we have worked through various use cases in small groups. The mixture between the consumption of examples and the use of one's own brainpower makes up the lessons with Steffen.

AI framework

The adapted canvas framework from Prolego is recommended to us for a clean project definition.

AI canvas from Prolego

The classic canvas according to Osterwalder & Pigneur places the values ​​on the left side of the model and the efficiency on the right side. Since AI projects often have to be integrated into an existing business, the definition is slightly adapted here. Here on the left side we have the business with:

  • Opportunity -> What is the profit in business due to the AI ​​solution?
  • Consumers -> Which stakeholders benefit? People and systems count as stakeholders.
  • Strategy -> How can I generate an advantage, internally or externally?
  • Policy & process -> Are there any ethical or legal issues to be considered?

On the right side, the technical feasibility is clarified in the AI ​​canvas:

  • Solution -> What does the technical solution look like? So which system architecture, models and workflows are necessary?
  • Data -> Where do we get our data from?
  • Transfer learning -> Where can I find existing knowledge about the solution I am looking for? Is there any research or even ready-made solutions?
  • Success criteria -> How do I measure the success of the model?

AI and ...

…Text

Typically, AI is used in the text area for search engines, the automatic classification or evaluation of texts, in translation or in article generation.

A rough distinction is made between NLP (Natural Language Processing), NLU (Natural Language Understanding) and NLG (Natural Language Generation).

  • NLP includes text type and word analysis as well as semantics and linguistics:
    • Example: perspective.ai uses a sentiment analysis to check text fragments, e.g. comments, whether they are negative, offensive, critical, etc.
  • The content analysis, i.e. the "understanding" of a text, is basically summarized under NLU:
    • Facebook does 4.5 billion translations every day. About NLU, you must first understand what the context is like before you can translate. A strict word-to-word translation would be absolutely unsatisfactory.
  • NLG automatically generates new text
    • In 2015, the New York Times built an AI that supplemented the authors' texts in real time with additional information from various sources.

…Image

Image recognition is certainly one of the prime disciplines for artificial intelligence. Contain images, which is of course logical for us humans:

  • Objects
  • Scenarios or places
  • Weather information
  • Date or time
  • Structure / texture

For example, it is possible to subsequently improve poor quality images or to colorize black and white images later - for example, to breathe new life into old photos. Colourise.sg does exactly that, you can try it yourself.

On the left the original photo, on the right the newly colored one

... video and audio

AI is used for analysis as well as for generating sounds or video. Well-known examples are certainly intelligent monitoring systems that recognize moving objects or even self-driving vehicles. It becomes dangerous with so-called deep fakes. It is therefore possible to create videos in which appearances by politicians are completely faked. In the future, moving images will also have to be critically checked for authenticity.

Example of a deep fake with Obama

An exciting example is Sunspring, a film based on a script written entirely by artificial intelligence. The script was filmed with real actors, you can find the result here.