Artificial

Intelligence:
Benefits and
Risks

PYQ-302 Seminar
Presentation
  From our Google assistant to the home automation
systems and self-driving cars, AI is the backbone of
this new age.

What is AI? By some, it is considered as the 'electricity' of



this developed growing world.
 AI is a simulation of human intelligence, a machine
(or a program) made to think and act like humans in
problem-solving and conversational skills, etc.
 Why the
fuss about AI?
Everyone talks about AI and Machine Learning. Main reason: AI can be applied in an extremely
wide range of areas. It is very versatile and
has improved, made profits for almost everyone
who uses it
  AI doesn’t get tired: 
• After all, AI is either  a machine or a software program. Neither of them can get
tired like humans do.
• They are much complex and more efficient than an average human, helping
them overcome the human limitations.
 It assists in large scale & small scale works AND keeps improving:

Benefits of AI • Most online/offline games that facilitate a 'bot' to play with a real human use AI
and they are very well trained to do it. For example: Playing chess with a bot
can be tougher than with a real human.
• They keep learning while working as well, for example the Google Photos
in the smartphone creates a new category of photos as soon as a new face
appears in a picture
• Assisting with GPS, Scheduling the week in the smartphone, and many more.
 One of the most important uses in large scale industries is the Fin-tech industries,
there AI and ML models are used for fraud detection, customer-ship based on past
customer behavior is detected. Eg: MasterCard, Visa, etc. [0]
  Medical applications [1]
• This is one of the most fruitful areas in which AI has brought a
revolution
• doctors and medical institutes use AI software to
Benefits •
detect presence of diseases, health related issues
Used in radiosurgery as well: which is very complicated and
(contd.) requires very high skilled staff to be performed without
incurring any damage to the surrounding cells and tissues.
• Clinical institutes use AI to counter-study the medication and
effects of diseases
  AI has made into the criminal justice system of some developed countries
(not all)

Breakthrough Why? Because AI works with complex models on facts and the data provided


from the past usage/experience, it doesn't involve emotions which can be an

obstacle in the crime investigation system. This removes the sense of bias

: AI in 
from the results.
It looks for patterns in criminal records and past data to recommend the

criminal
chances of a person being the culprit.
 But, it has been noticed that some innocent people have been found guilty by
the AI programs (also called the misclassifications in tech jargon), and

justice 
thus physical human investigation is still required because of this issue. [2]
One very good benefit is that prediction can be quick, instead of waiting for
multiple days to identify the criminals, AI can do it in a very little time. But
again, it is useless if the prediction isn't perfectly correct.
 Invasion of privacy [3]
• As heard in the news and social media posts on many occasions that Amazon Alexa, Google
Home tend to listen to regular conversations at home in order to improve the device and to make
it more compatible with the users, there's a threat of privacy invasion lurking.
• Social media ads suggestions and google search suggestions tend to be similar to what the
person has been talking about, gives the idea that the privacy is at stake, if the data is breached
then it can be used by some people for wrong purposes (like, credit card details)

Risks For a better understanding of how this can be problematic, let's consider the case of


China, where the social credit system was implemented:

• Notes minor offenses like smoking in non-smoking zones (I.e., tracking the location of the
person)
• Keeping track of the financial records
• Taxes
• Private messages too
• Media usage
 These breaches can be very problematic if this data is acquired by some third party like
a hacker who can use these sensitive information for wrong purposes.
 Problems with no emotions:
 On one hand, this attribute of AI can be beneficial (like in the criminal
justice system), on the other, it is an obstacle in the development of AI
 ML and NLP seem to be a mini step towards this development which is

Risks (contd.)
used in chatbots, still it requires human interventions at times to resolve
some issue or to handle the user's problems with some emotions.
 If the whole of it is left to bots, customer experience across the globe
will go downhill.
 Sometimes, a bot cannot understand your pain point because you do not
emotionally drive it. You will always need human ears to get things
done.
  Lack of qualified professionals: 
• Another snag that frequently happens at the business level for
the appropriation of AI is the shortage of profiles with abilities
and involvement with this kind of executions. It is pivotal in
these cases to have experts who have effectively chipped away
at activities of the equivalent magnitude.

Risks (contd.)  Data availability: 

• Regularly, information is introduced in confinement across
organizations or is conflicting and of inferior quality,
introducing a critical test for organizations looking to make
esteem from AI at scale. To conquer this boundary, it will be
indispensable to draw up a reasonable technique from the
beginning with the goal that AI information can be separated in
a coordinated and reliable manner. 
 Conclusion

 Many individuals fear AI turning 'mindful' and wiping out humanity. For
quite a long time, many sci-fi medias like movies, articles, blogs have
protrayed the same. 
 The fantasies about AI turning out to be evil aren't completely right. 
 While the danger is genuine, there are defends set up to keep an AI end
of the world from occurring.
 We're many years from a hyper savvy machine, and until further notice,
the advantages of AI far exceed the dangers implied.
  [0]: 
https://www.mastercard.us/en-us/business/issuers/business-p
ayments/fraud-prevention.html

References  [1]: 
https://www.quantib.com/the-ultimate-guide-to-ai-in-radiolo
gy
 [2]: 
https://www.nytimes.com/2020/06/24/technology/facial-reco
gnition-arrest.html
 [3]: 
https://mytechdecisions.com/compliance/is-artificial-intellig
ence-data-collection-an-invasion-of-privacy/