ML AI ETHICS
ML AI ETHICS
Preexisting Biases:
if police had not registered complaints on particular segment of society and this data-set trains an Al
system then it will tend to give lower priority index while actual they should be treated with equality.
Technical biases: Such biases may creep into the system consciously or unconsciously by the
flaw of logic. A reservation system may tend to slow down due to higher number of request
from a particular region thereby making everyone wait longer than usual to book a seat
Emergent biases: These biases occur when system comes in use. For example, a user interface system for
fixing doctor's appointment is too difficult to use by physically challenged persons.
Security and Privacy:
Think Of an Al-system that could devise ways to hack into user accounts after getting trained with the rich
data-sets. Possibilities of false alarms, auto- locking due to logic malfunction are very much there. Issues
related with cybersecurity, cyberbullying and data-theft need to be addressed with different perspectives.
Decision making, common sense and human values: Harms caused due to unjust decisions, uncalled for
actions and lack of decision-making capability in a learning machine may create more problems than the
solution it provides. A sinking ship, where Al system locks the door preventing water to flood the rooms
while the other system tends to open the same doors due to rising smoke. How Al-systems will imitate
common sense and take care of human dignity and values.
AI Access:
AI may be not accessible by everyone. The sections of society who can afford and who have access to the
technology and devices can enjoy the benefits of the power of Al but many sections of the society may not
have that privilege.
Access to Al for Developers:
As an emerging technology, the access to the training for Al and to the Al development tools is not available
to everyone. Very selective enterprises have that privilege and greater initiatives are needed to let young
people have access to the facilities to get trained in Al concepts.
Access to Al for Rural Areas:
How common man in rural regions should be able to access the Al enabled services in future should be
considered. Al has very impactful applications in agriculture.
Access to Al by Disabled
Al enabled services should support features which enable physically challenged people to operate them.
Affordable Access for All
Al infrastructure and training facilities should be available as part of public services for all sections of
society.
AI and Employment:
Al is a promising technology for sustainable growth for everyone, but it may affect the people where
human labour and middle level skills are required. People who do not upgrade themselves will be quickly
replaced by Al enabled system. This will lead to the major problem of increase in unemployment among
non-skilled, semi-skilled and traditionally skilled workers.
Al should possibly be dealt as a necessary evil by striking balance between its benefits and losses keeping
common masses in mind.
Preexisting Biases:
if police had not registered complaints on particular segment of society and this data-set trains an Al
system then it will tend to give lower priority index while actual they should be treated with equality.
Technical biases: Such biases may creep into the system consciously or unconsciously by the
flaw of logic. A reservation system may tend to slow down due to higher number of request
from a particular region thereby making everyone wait longer than usual to book a seat
Emergent biases: These biases occur when system comes in use. For example, a user interface system for
fixing doctor's appointment is too difficult to use by physically challenged persons.
Security and Privacy:
Think Of an Al-system that could devise ways to hack into user accounts after getting trained with the rich
data-sets. Possibilities of false alarms, auto- locking due to logic malfunction are very much there. Issues
related with cybersecurity, cyberbullying and data-theft need to be addressed with different perspectives.
Decision making, common sense and human values: Harms caused due to unjust decisions, uncalled for
actions and lack of decision-making capability in a learning machine may create more problems than the
solution it provides. A sinking ship, where Al system locks the door preventing water to flood the rooms
while the other system tends to open the same doors due to rising smoke. How Al-systems will imitate
common sense and take care of human dignity and values.
AI Access:
AI may be not accessible by everyone. The sections of society who can afford and who have access to the
technology and devices can enjoy the benefits of the power of Al but many sections of the society may not
have that privilege.
Access to Al for Developers:
As an emerging technology, the access to the training for Al and to the Al development tools is not available
to everyone. Very selective enterprises have that privilege and greater initiatives are needed to let young
people have access to the facilities to get trained in Al concepts.
Access to Al for Rural Areas:
How common man in rural regions should be able to access the Al enabled services in future should be
considered. Al has very impactful applications in agriculture.
Access to Al by Disabled
Al enabled services should support features which enable physically challenged people to operate them.
Affordable Access for All
Al infrastructure and training facilities should be available as part of public services for all sections of
society.
AI and Employment:
Al is a promising technology for sustainable growth for everyone, but it may affect the people where
human labour and middle level skills are required. People who do not upgrade themselves will be quickly
replaced by Al enabled system. This will lead to the major problem of increase in unemployment among
non-skilled, semi-skilled and traditionally skilled workers.
Al should possibly be dealt as a necessary evil by striking balance between its benefits and losses keeping
common masses in mind.