Artificial intelligence is a technology that is already
impacting how users interact with, and are affected by the Internet. In the
near future, its impact is likely to only continue to grow. AI has the
potential to vastly change the way that humans interact, not only with the
digital world, but also with each other, through their work and through other
socioeconomic institutions – for better or for worse.
Artificial intelligence (AI) traditionally refers to an
artificial creation of human-like intelligence that can learn, reason, plan,
perceive, or process natural language.
Artificial
intelligence (AI) has received increased attention in recent years. Innovation,
made possible through the Internet, has brought AI closer to our everyday
lives. These advances, alongside interest in the technology’s potential
socio-economic and ethical impacts, brings AI to the forefront of many
contemporary debates. Industry investments in AI are rapidly increasing, and
governments are trying to understand what the technology could mean for their
citizens.
The
collection of “Big Data” and the expansion of the Internet of Things (IoT), has
made a perfect environment for new AI applications and services to grow.
Applications based on AI are already visible in healthcare diagnostics,
targeted treatment, transportation, public safety, service robots, education
and entertainment, but will be applied in more fields in the coming years.
Together with the Internet, AI changes the way we experience the world and has
the potential to be a new engine for economic growth.
Few
Uses of AI
Although
artificial intelligence evokes thoughts of science fiction, artificial
intelligence already has many uses today, for example
• Email filtering: Email services use
artificial intelligence to filter incoming emails. Users can train their spam
filters by marking emails as “spam”.
• Personalization: Online services use
artificial intelligence to personalize your experience. Services, like Amazon
or Netflix, “learn” from your previous purchases and the purchases of other
users in order to recommend relevant content for you.
• Fraud detection: Banks use artificial
intelligence to determine if there is strange activity on your account.
Unexpected activity, such as foreign transactions, could be flagged by the
algorithm.
• Speech recognition: Applications use
artificial intelligence to optimize speech recognition functions. Examples
include intelligent personal assistants, e.g. Amazon’s “Alexa” or Apple’s
“Siri”.
Artificial
intelligence is further defined as “narrow AI” or “general AI”. Narrow AI,
which we interact with today, is designed to perform specific tasks within a
domain (e.g. language translation). General AI is hypothetical and not domain
specific, but can learn and perform tasks anywhere. This is outside the scope
of this paper. This paper focuses on advances in narrow AI, particularly on the
development of new algorithms and models in a field of computer science
referred to as machine
learning.
Challenges
Decision-making: transparency and “interpretability”. With artificial
intelligence performing tasks ranging from self-driving cars to managing
insurance payouts, it’s critical we understand decisions made by an AI agent.
But transparency around algorithmic decisions is sometimes limited by things
like corporate or state secrecy or technical literacy. Machine learning further
complicates this since the internal decision logic of the model is not always
understandable, even for the programmer
Data Quality and Bias. In machine
learning, the model’s algorithm will only be as good as the data it trains on –
commonly described as “garbage in, garbage out”. This means biased data will
result in biased decisions. For example, algorithms performing “risk
assessments” are in use by some legal jurisdictions in the United States to
determine an offenders risk of committing a crime in the future. If these
algorithms are trained on racially biased data, they may assign greater risk to
individuals of a certain race over others. Reliable data is critical, but
greater demand for training data encourages data collection. This, combined
with AI’s ability to identify new patterns or re-identify anonymized
information, may pose a risk to users’ fundamental rights as it makes it
possible for new types of advanced profiling, possibly discriminating against
particular individuals or groups.
Safety and Security. As the AI agent
learns and interacts with its environment, there are many challenges related to
its safe deployment. They can stem from unpredictable and harmful behavior,
including indifference to the impact of its actions. One example is the risk of
“reward hacking” where the AI agent finds a way of doing something that might
make it easier to reach the goal, but does not correspond with the designer’s
intent, such as a cleaning robot sweeping dirt under a carpet.
Social and Economic Impact. It is predicted
that AI technologies will bring economic changes through increases in
productivity. This includes machines being able to perform new tasks, such as
self-driving cars, advanced robots or smart assistants to support people in
their daily lives. Yet how the benefits from the technology are
distributed, along with the actions taken by stakeholders, will create vastly
different outcomes for labor markets and society as a whole.
Governance. The institutions, processes and
organizations involved in the governance of AI are still in the early stages.
To a great extent, the ecosystem overlaps with subjects related to Internet
governance and policy. Privacy and data laws are one example.
Guiding Principles
and Recommendations:
- Adopt
ethical standards: Adherence to the principles and standards of ethical
considerations in the design of artificial
intelligence, should guide researchers and industry going
forward.
- Promote
ethical considerations in innovation policies:
Innovation policies should require adherence to ethical standards as a
pre-requisite for things like funding.
- Ensure
Human Interpretability of Algorithmic Decisions: AI
systems must be designed with the minimum requirement that the designer
can account for an AI agent’s behaviors. Some systems with potentially
severe implications for public safety should also have the functionality
to provide information in the event of an accident.
- Empower
Users: Providers of services that utilize AI need to
incorporate the ability for the user to request and receive basic
explanations as to why a decision was made.
- “Algorithmic
Literacy” must be a basic skill: Whether it is the curating of
information in social media platforms or self-driving cars, users need to
be aware and have a basic understanding of the role of algorithms and
autonomous decision-making. Such skills will also be important in shaping
societal norms around the use of the technology. For example, identifying
decisions that may not be suitable to delegate to an AI.
- Provide
the public with information: While full transparency
around a service’s machine learning techniques and training data is
generally not advisable due to the security risk, the public should be
provided with enough information to make it possible for people to
question its outcomes.
- Humans
must be in control: Any autonomous system must allow for a human to
interrupt an activity or shutdown the system (an “off-switch”). There may
also be a need to incorporate human checks on new decision-making
strategies in AI system design, especially where the risk to human life
and safety is great.
- Make
safety a priority: Any deployment of an autonomous system should be
extensively tested beforehand to ensure the AI agent’s safe interaction
with its environment (digital or physical) and that it functions as
intended. Autonomous systems should be monitored while in operation, and updated
or corrected as needed.
- Privacy
is key: AI systems must be data responsible. They should use
only what they need and delete it when it is no longer needed (“data
minimization”). They should encrypt data in transit and at rest, and
restrict access to authorized persons (“access control”). AI systems
should only collect, use, share and store data in accordance with privacy
and personal data laws and best practices.
- Think
before you act: Careful thought should be given to the instructions
and data provided to AI systems. AI systems should not be trained with
data that is biased, inaccurate, incomplete or misleading.
- If
they are connected, they must be secured: AI systems that are
connected to the Internet should be secured not only for their protection,
but also to protect the Internet from malfunctioning or malware-infected
AI systems that could become the next-generation of botnets. High
standards of device, system and network security should be applied.
- Responsible
disclosure: Security researchers acting in good faith should be
able to responsibly test the security of AI systems without fear of
prosecution or other legal action. At the same time, researchers and
others who discover security vulnerabilities or other design flaws should
responsibly disclose their findings to those who are in the best position
to fix the problem.
- Ensure
legal certainty: Governments should ensure legal certainty on how
existing laws and policies apply to algorithmic decision-making and the
use of autonomous systems to ensure a predictable legal environment. This
includes working with experts from all disciplines to identify potential
gaps and run legal scenarios. Similarly, those designing and using AI
should be in compliance with existing legal frameworks.
- Put
users first: Policymakers need to ensure that any laws applicable
to AI systems and their use put users’ interests at the center. This must
include the ability for users to challenge autonomous decisions that
adversely affect their interests.
- Assign
liability up-front: Governments working with all stakeholders need to
make some difficult decisions now about who will be liable in the event
that something goes wrong with an AI system, and how any harm suffered
will be remedied.
- Social and Economic
Impacts: All stakeholders should
engage in an ongoing dialogue to determine the strategies needed to seize
upon artificial intelligence’s vast socio-economic opportunities for all,
while mitigating its potential negative impacts. A dialogue could address
related issues such as educational reform, universal income, and a review
of social services.
- Promote
Multistakeholder Governance:
Organizations, institutions and processes related to the governance of AI
need to adopt an open, transparent and inclusive approach. It should be
based on four key attributes: Inclusiveness
and transparency; Collective responsibility; Effective decision making and
implementation and Collaboration through distributed and interoperable
governance
Thanks to the Source with References
Notes
Images Sources : Google, Internet