Strategic Approaches to AI Flashcards
what has AI generated?
excitement
scepticism
concern
key questions regarding AI?
what’s an AI project?
what roles are involved?
what skills do they need?
how do people & roles connect to AI systems?
what are the differences between AI systems and AI management systems?
AI systems vs AI management systems?
ai systems = tools and processes that utilise AI to produce their results/output
ai management systems = mechanisms that seek to ensure the ethical use of AI technologies and direct how the tools which incorporate AI should be managed, evaluated & improved
what must AI policies do?
should set out a framework for the organisation’s AI objectives and be suitable for that particular organisation
must describe how systems/components will be monitored/reviewed so deficiencies are addressed
AI Builders?
AI deployers?
ai builders = those involved in approving, designing, constructing and maintaining AI systems
ai deployers = members of an organisation who use AI systems internally or in support of external clients/stakeholders
ai stakeholders?
management - make decisions
data scientists - construct AI tools
staff - employ the tools
clients - interact with the AI tools
key demands which must be met by management when implementing AI systems?
- comprehensive & robust AI policy
- sufficient resources
- assessed impacts
- communication of reasons for implementation
- ensuring that human resources are sufficient
- emphasising the importance of design
- carrying out other tasks
factors that affect how an ai system is designed/implemented
- organisation & structure
- industry
- culture
- resources
- customer attitudes
ai system groups?
service providers
partner organisations
software developers
customers
regulators
key task of ai management system
determine the objectives
objectives of ai tools
measurable
incorporate requirements set out
plan for how objectives are to be achieved
ai management system should commit to following system:
shouldn’t cause harm
achieve it’s objectives
able to undergo continuous improvement
management will design ai risk assessment process?
yes
risk treatment process?
risk treatment plan?
process = what should be used when risks have been assessed
plan = maps out what management will do to manage/treat risk
impact assessment
determining the effects & impacts of adopting ai systems
takes into account wellbeing of individuals
what to consider when creating an impact assessment?
- situations where impact assessment’s performed
- complexity of AI system
- robustness or AI system
- precise steps contained in assessment
- staff assigned
- how it’s used for continuous improvement of ai system
planning process
identifying risks (risk assessments)
incorporate steps to benefit from opportunities to mitigate risks
determine the scope of the AI system
key ethical requirements of an AI system?
predictability
availability
maintainability
framework for making choices
DIODE
define questions
issues analysis
options evaluation
decision determination
explanation dissemination
risks when ethical requirements aren’t properly ensured
reliability & safety
privacy & security
fairness
building & deploying an AI management system isn’t and end point, its the beginning of a process including
monitoring
evaluation
continuous improvement
do management need to determine the communication policy?
yes
other key aspects management need to address?
- rules of distribution of documents
- control of access within the organisation
- rules on the authority to make changes
- length of time & security of document storage
what falls under the support function?
protection of the data used in an AI system and reliable storage of all associated documentation falls under support function
