HUMAN RIGHTS IN THE AGE OF ARTIFICIAL INTELLIGENCE
7. ALGORITHMIC DECISION-MAKING: Using outputs produced by algorithms to make decisions. One
of the earliest forms of algorithmic decision-making that is still in use today in the United States is federal
sentencing guidelines for judges. This involves nothing more than a weighted mathematical equation,
drawn from statistics, that recommends a sentence length based on the attributes of the crime.
19
8. ROBOT: Robots often use many of the forms of artificial intelligence described above. However, by
definition, robots have a physical body and mobility. Robots that use AI are able to perceive changes in
their environment and function accordingly.
20
Although robots typically come to mind when thinking
about artificial intelligence, they currently constitute a very small amount of our interactions with AI. AI
in the field of robotics is a growing area of research and development but has not yet made nearly as many
advancements or become as ubiquitous as non-robotic forms of machine learning.
21
9. BOTS: Software applications that run automated tasks. Bots are increasingly being powered by ML,
particularly chatbots, which use natural language processing to conduct human-like conversations with users.
10. OPEN DATA: Data that is freely available for everyone to view, use, share, and re-publish without
restrictions. There is a broad open data movement that advocates that data should generally be treated
this way.
22
In the context of AI, many advocates suggest training data for ML systems be open in order to
surface bias and errors, as well as to shed light on the outputs ML systems produce. There is controversy on
the best method to do this while respecting the privacy interests of data subjects.
11. PROTECTED INFORMATION: Information that includes, reflects, arises from, or is about a person’s
communications, and that is not readily available and easily accessible to the general public. While it has long
been agreed that communications content deserves significant protection in law because of its capability
to reveal sensitive information, it is now clear that other information arising from communications—
metadata and other forms of non-content data—may reveal even more about an individual than the content
itself, and thus deserves equivalent protection. Today, each of these types of information might, taken
alone or analyzed collectively, reveal a person’s identity, behavior, associations, physical or medical
conditions, race, color, sexual orientation, national origins, or viewpoints; or enable the mapping of the
person’s location, movements or interactions over time, or of all people in a given location, including around
a public demonstration or other political event.
23
Do'stlaringiz bilan baham: |