Page 11 - AI Vol 2: Risks of AI
P. 11
results, whereas some tasks are just too complex
for the current AI system to provide effective WHEN EMPLOYEES RELY
assistance. Continually monitoring and evaluating HEAVILY ON AI TOOLS, THEY
AI performance ensures that the agency can adapt MAY EXPERIENCE A DECLINE IN
their use of AI systems (or evaluate whether other CRITICAL SKILLS, PARTICULARLY
systems are more appropriate). In addition to the IN AREAS REQUIRING COMPLEX
general liability risks discussed in this document, ANALYSIS AND DECISION-MAKING.
an agency which deploys but fails to appropriately
monitor an AI system could be found liable under overreliance. In addition to this situation leading
a negligence theory. This liability may arise if to poor decision making and performance of the
the AI model generates harmful or inappropriate public agency, it may also lead to legal liability
responses, whether directed towards a member of to the extent that the agency implements flawed,
the public, a student, or an employee relying on the biased, or legally non-compliant decisions based
AI system for decision-making. Such liability could on AI advice. It is thus critical for agencies to
be attributed to the agency's failure to oversee the implement protocols to guard against these risks.
deployment adequately, especially if the potential
risks were foreseeable with proper monitoring.
AI system stability and long term sustainability
is also important to consider, particularly in the
Related issues include overreliance and the context of overreliance and skill degradation.
subsequent degradation of human skills. While still in the infancy of widespread AI
Overreliance refers to situations where AI systems development and deployment, there are a large
are excessively depended upon for decision- number of companies attempting to build and
making, potentially leading to diminished human develop businesses and AI products, some of
engagement and a failure to critically assess which may not remain viable over time. Over
AI outputs. This reliance can result in a lack of the Thanksgiving break in 2023, OpenAI saw
questioning of AI-generated results, allowing its CEO fired and the vast majority of its staff
unchallenged errors or biases to influence threatened to quit in protest, creating concern
decisions. Moreover, an important concern among the business built upon or simply relying
that arises with the overuse of AI is the risk of on OpenAI’s services. While that situation was
skill degradation among staff. When employees ultimately resolved without interruption of
rely heavily on AI tools, they may experience OpenAI’s services, the episode illustrates that
a decline in critical skills, particularly in areas agencies need to think critically regarding the
requiring complex analysis and decision-making. extent they intend to rely on any one AI service
This degradation not only diminishes individual and have backup plans in the event a service is
capabilities but also impacts the overall resilience no longer available. Boardroom conflicts are just
and adaptability of the agency. These issues are one of the potential events that could lead to the
very related, as overreliance can lead to skill discontinuation of AI services. AI models are
degradation, and skill degradation can lead to
trained on vast amounts of information primarily
RISKS OF AI | LOZANOSMITH.COM VOLUME 2 | 11