This video belongs to the openHPI course Künstliche Intelligenz und maschinelles Lernen für Einsteiger. Do you want to see more?
An error occurred while loading the video player, or it takes a long time to initialize. You can try clearing your browser cache. Please try again later and contact the helpdesk if the problem persists.
Scroll to current position
- 00:00In this unit we want to look at the general limitations of machine learning and artificial intelligence, before we go into individual aspects in detail.
- 00:08In general, the following risks and problems exist on the topic of machine learning or artificial intelligence.
- 00:16In general, there are errors that can cause the Data Scientists or machine learning experts can pass.
- 00:22Potential for manipulation, misuse of data or problems, which result from a data monopoly situation.
- 00:29The subitem Errors mainly refers to actual errors made by people, who work on machine learning models.
- 00:36This involves issues such as spurious correlations, general bias or lack of objectivity in data analysis and for example the Simpson paradox.
- 00:46With the Simpson paradox, for example, it seems, that the evaluation of different groups is different, depending on whether you combine the results of the groups or not.
- 00:58If you do not pay attention to this, you can draw erroneous conclusions.
- 01:02Manipulation refers mainly to the changes in results, to better represent the performance of the model.
- 01:10For example, you can adapt your training to exactly those data, on which this is tested.
- 01:17So the model seems to be super, but this is only the case for exactly these data.
- 01:22But the model does not generalize at all against unseen new data points.
- 01:27Of course, machine learning also carries the risk of being misused, either there is discrimination in data or one can discriminate through data
- 01:38Monitoring and data breaches are also important and critical issues
- 01:44Of course, large amounts of data and the associated value problems or social issues.
- 01:53Is all data limited to only a few players such as a few companies, then these companies certainly have some power, because data can be seen as a kind of resource.
- 02:05This often makes equal opportunities and aspects such as control difficult.
- 02:10Aspects from these problem areas we will in the different videos still exemplary regard.
- 02:19In general the following problems exist: The lack of knowledge as well as the lack of ethics and regulation.
- 02:28The general solution is regulation, education and ethical guidelines, and a very general social discussion,
- 02:37We will be working in the following units to some of the aspects just mentioned,
- 02:45the importance of data and the limitations it imposes with the question what we really know, with the topic of discrimination
- 02:54and also with the topic of ethics and morals in the context of machine learning.
To enable the transcript, please select a language in the video player settings menu.