academic approach, to risk mitigation, 157–158
Accenture, 124
accuracy rates, 76
AI. See artificial intelligence (AI)
AI agents, 103
AI assistants
human traits in, 100
Aiden, 209
third-party, 66
AI systems
data-centric approach to building, 40–43
dependability of, 63
employee involvement in design of, 131, 133
flexibility of, 64
production and deployment of, 39–40, 42, 43
scalability and extendibility of, 64–65
training of, by humans, 99–100
Albert, 121
Alexa, 99
algorithms, 140
consumer reaction to effects of, 172–173
deployment of, 168
detrimental effects of, 168–170
market conditions and, 168, 170
perception of, by targeted users, 170–172
Amelia, 32
amplification, of human capabilities, 102–103
Andreessen, Marc, 195
artificial intelligence (AI)
See also machine learning
human collaboration with, 97–116
AT&T, 115
automation
of processes, 208
using AI, 98
automation projects
AutoNLP, 200
autonomous systems, 140, 144, 145
Baidu, 37
in algorithms, 151, 155–156, 161, 167–170
detection of, 161
sources of, 185
black-box problem, 33, 100, 121, 136, 182–183
Blackman, Reid, 155–166, 179–186
Borealis AI, 209
business processes
redesigning for collaborative intelligence, 98–116
for reinforcement learning, 210–211
Campbell, Craig, 80
capabilities, assessment of existing, 91
Carnival Corporation, 111, 114
Center of Excellence (COE) model, 220–221
clustering, 22
cocreation, 105
cognition, theory of distributed, 134
collaborative intelligence, 97–116
collective intelligence, 134–136
confidence rates, 74
consumer internet companies, 37–40
consumer reactions, to AI algorithms, 172–173
control, loss of, 119, 130–131
corporate culture, 51
counterfactual explanations, 136
credit approval, 101
customer interactions, 103–104
customer service, 31–33, 103–104
customization
human-machine collaboration for, 106–107
DALL·E 2, 199
Danks, David, 135
Danske Bank, 108
data
biased, 185
bottlenecks, 63
curation, 30
enterprise data strategy, 93
image, 50
input, 85
for no-code platforms, 192–193
overfitting, 22
personal, 102
separating signal from noise in, 20–24
visualization, 49
data-centric AI development, 40–43
data compliance officers, 101
data governance board, 159–160
data scientists, 33–34, 51, 53, 55, 71, 218, 222
Davenport, Thomas H., 27–35, 55
decision-making
analytics for, 49
deep learning and, 33
explanations for, 100–101, 136, 163, 182–183
human in the loop (HITL), 141–142
human in the loop for exceptions (HITLFE), 142–143
human on the loop (HOTL), 143–144
human out of the loop (HOOTL), 144
in uncertainty, 80
using collaborative intelligence, 110–111
using wide data, 19
decision-making tools, 139–145
deep learning, 30, 32, 33, 48, 50, 52
design thinking, 5
See also biases
distributed cognition, 134
Drucker, Peter, 131
dystopians, 112
embodiment, of AI, 104
employees
adoption of AI and, 117–122, 123–137
fear of being replaced by AI, 123–124
impact of AI on morale of, 225–236
incentivizing to identify AI ethical risks, 164
negative impacts of AI on, 133–134
new roles and skills for, 114–115
resistance to change by, 119–120
employment opportunities, 101
endowment effect, 172
ethical issues, 34, 101–102, 155–166, 223
defining ethical AI standards, 182–183
high-level AI ethics principles for, 159
“on-the-ground” approach to, 158
organizational awareness of, 163–164
ethical risk framework, 160–161
ethics managers, 101
exceptions, in decision-making, 142–143
exoskeletons, 104
expert systems, 33
explanations
See also black-box problem
for AI decisions, 100–101, 163, 182–183
counterfactual, 136
meaning of, 136
extended mind, 134
face recognition, 21
failures, 51
plan for dealing with, 147–151
false negatives, 194
false positives, 194
Fast Forward Labs, 47
Feature Stores, 219
foundational models, 199, 202–203
fraud detection, 107–108, 130, 194
gap analysis, 91
general artificial intelligence, 202–203
General Data Protection Regulation (GDPR), 100–101
GitHub, 199
Google, 37, 126, 156, 159, 199, 206, 209
governance teams, 159–160, 218, 222, 223
Harmer, Peter, 5
health care, ethics in, 161–162
health treatment recommendations, 31
home security alarms, using AI Canvas for, 80–87
Hugging Face, 200
human in the loop (HITL), 141–142
human in the loop for exceptions (HITLFE), 142–143
human learning, 14
human on the loop (HOTL), 143–144
human out of the loop (HOOTL), 144
humans
assistance of machines by, 99–102
collaboration between AI and, 97–116
machines assisting, 102–104, 124
Hutchins, Edwin, 134
Hyundai, 104
image data, 50
implementation phases, 125–137
implicit bias, 185
informed consent, 162
intelligence amplification, 141–142
intelligent machines, as “colleagues,” 4–5
interoperability, 222
investment decisions, 132
job losses, 34, 112–113, 123–124, 133–134
job opportunities, 101
judgmental bootstrapping, 126–127
Knickrehm, Mark, 115
knowledge work, automation of, 90
Koko, 100
of AI, 112–113, 123–124, 133–134
of language-based AI, 201
language-based AI tools, 197–204
large language models, 198–199
lead scoring, 191
legacy industries, use of AI in, 37–40
LIME, 183
linear regression, 73
machine learning
See also artificial intelligence (AI)
applications of, 13–15, 18–20, 28–30, 32–33
big data and, 18
cross-validation and, 20, 23–24
feature extraction and, 20, 21–22
mistakes to avoid using, 24–25
predictive analytics and, 55–59
supervised, 20, 72–77, 207–208
understanding, 13–16, 17–25, 33–34
unsupervised, 22
machine learning operations (MLOps), 40–43, 218
management options, for micro-decisions, 141–145
managers
creativity needed by, 5
knowledge of machine learning by, 17–25
time spent on administrative tasks by, 2–3
Marble Bar Asset Management (MBAM), 126, 129, 132, 133
Martinho-Truswell, Emma, 11–16
maturity models, 90
Mayflower Autonomous Ship, 144
medical care prediction algorithm, 167–168
medium-size businesses, AI for, 189–195
MLOps. See machine learning operations
Model Catalogs, 219
model-centric development, 40–41
natural language processing (NLP), 126, 197–204
preparing for future of, 200–203
“on-the-ground” approach, to risk mitigation, 158
optimistic realists, 113
optimization, of processes, 208
organizational awareness, of ethical issues, 163–164
outcomes, of actions, 85
overfitting, 22
Pandora, 111
personal data, 102
personalized recommendations, 19
false results in, 194
lowering cost of, 80
predictive analytics, 53–59, 73
predictive models, 55
preferences, 19
privacy issues, 102, 130, 133, 135, 137, 161–162
limitations of AI, 15
process automation, 27–29, 33, 208
See also business processes
bespoke, 219
for building and operationalizing AI models, 218–219
productivity skeptics, 113
product managers, ethical guidance for, 162–163
project opportunities
projects, sequencing, 91
proof of concept, 39–40, 42, 43
RAID (Research Analysis and Information Database), 126
reasoning capabilities, 128
recidivism prediction algorithm, 167
recommendation systems, 31, 49, 50, 125–126
regression analysis, 52
regulated industries, “black-box” issue in, 33
reinforcement learning, 205–213
applications of, 206–207, 209–210
relationships, disruption of, 120
report writing, 3
research and development (R&D), 49, 51
resistance, to AI, 119–120, 123
high-level AI ethics principles for, 159
“on-the-ground” approach to, 158
robotic process automation (RPA), 27–29, 33
accidents caused by, 148
Royal Bank of Canada, 209
rule-based expert systems, 33
safety engineers, 101
sales prospects, 191
standardization of model building and, 218–219
Schlesinger, Leonard A., 225–236
Schmidt, Eric, 203
search algorithms, 49
Sedol, Lee, 205
selection, 23
self-determination, 161
self-driving cars, 121–122, 140, 182
sequential decision tasks, 206–207, 209–210
SHAP, 183
shrinkage, 23
Siri, 99
small businesses, AI for, 189–195
smart-pricing algorithm, 168–173
accidents, 148
software-centric development, 40–43
software development, 201
spam filters, 49
Spotify, 206
stakeholders
engaging, in ethical issues, 164–165
multiple, 223
standardization, of AI model building, 218–219
Starbucks, 111
supervised learning, 20, 72–77, 207–208
sympathy, 100
talent recruitment, 34
tasks
breaking down, 75
teams
assessing existing capabilities of, 91
capability building of, 90, 93
for ethics risk mitigation, 180–181
governance, 159–160, 218, 222, 223
technology optimists, 113
text analytics, 200
text summarization, 50
theory of distributed cognition, 134
Thompson, Layne, 4
training, of machines by humans, 99–100
unconscious bias, 185
See also bias
understanding, trust and, 135
unsupervised learning, 22
utopians, 112
Vanguard, 32
Verneek, 201
wearable robotic devices, 104
West, Tessa, 131
Wilson, Andrew, 124
worker displacement, 34, 112–113, 123–124, 133–134