CANCELLED: AI - Machine Learning

Despite the name, artificial intelligence is not intelligent at all - but it can behave in a way that appears intelligent. Artificial intelligence (or machine learning) is nothing but computer programs that are designed to solve a specific task.

Despite the name, artificial intelligence is not intelligent at all - but it can behave in a way that appears intelligent. Artificial intelligence (or, to be more precise, machine learning) is nothing but computer programs that are designed to solve a specific task. It that sense, they are similar in scope to more traditional programs, where the computer solves a task by following a set of rules written by humans (e.g. "when the user clicks on a link, redirect the user to the web address it points to"). Contrary to traditional programs, however, machine learning does not need a predefined set of rules telling it how to solve a task. Instead, you tell the program what you want to solve, and the program will then figure out how to achieve that by itself. It does that by acquiring experience from huge amounts of data and from this learns how to solve the task. This usually requires large and complex machine learning models where neural networks are among the most famous - or infamous. Today often referred to simply as "deep learning", they work incredibly well and are the foundation in anything from Facebook's face recognition over Google Translate to autonomous cars, but there is a big problem: they are "black boxes". That is, it is extremely difficult to understand what they have learnt and why they make certain decisions - and this is a serious problem if one wishes to use them in, say, health care or the criminal justice system. In fact, they are so complex that we still don't understand why they actually work. But we can get a bit of intuition about how they work, and that is something we will look into in this talk.

 

In the talk, we will start with a conceptual introduction to machine learning to get an idea about what makes the field so unique and which types of tasks it can and cannot solve. From this, we will take a deeper look at neural networks - the, by far, most popular method in machine learning - and try to gain an intuition about how they work. Lastly, we will turn to our every day - we will see how far we have come with the development of artificial intelligence and discuss the impact this new technology will have on society.

Information
  • When

    25. okt. 2018 17:30 - 19:00
  • Where

    Biocenter, KU, Ole Maaløes Vej 5, 2200 København N

  • Registration Deadline

    24. okt. 2018 - 23:55

  • Organizer

    Kurser for studerende

  • Available Seats

    97

  • Event Number

    328841