Why high-risk AI systems may be a problem.
The forthcoming EU Artificial Intelligence Act (EU AIA) identifies the following application areas as high-risk: biometric identification; law enforcement; immigration and border control; criminal justice; among others. These are particularly important because they require completely accurate identification of people. While this is difficult for skilled professionals to do, the problem with artificial intelligence is that the users of AI tools for such purposes may not be able to check how an AI application reaches a decision about identification. (The AI application may not be transparent or explicable.) Users of AI applications in these areas must place some trust in the AI identification decision with limited knowledge. This is why these areas have been classified as high-risk.
AI image recognition has been demonstrated as useful for the location of individuals of endangered species, but we humans are not endangered. Any data source that was used for the above high-risk applications regarding people might include many people, all looking similar. Technical, standards and assessment progress is needed for AI applications in these high-risk areas, before they can be confidently deployed.
Endangered species encounters human photographer
At Coevolve IT we have the expertise to help you understand and use AI effectively. Get in touch!