Index
A
Abuse
approaches
abuser out of loop
add costs
less interesting for abusers
machine learning, adversary
business
boosting content
computers, compromising
direct theft
driving traffic
personal information, compromising
scales
SeeScales, abuse
suppressing content
telemetry
ways affecting intelligent system
Anti-malware system
APIs
Automated experiences
B
Balance intelligent experiences
cost of mistakes
factors
forcefulness
frequency
intelligence quality
value of success
Betting-based-interaction
Big problems
C
Changes in intelligence
Checklist, intelligent system project
approach
creation
implementation
orchestration
planning
Conceptual machine learning
algorithm
process
Confidence interval
Content processing
Contexts
forms
intelligence creation
runtime intelligence
Cost of implementation
Curves
D
Data
evaluation of intelligence
historical data, risks
partition data
by identity
by time
test set
Data experience
betting-based-interaction
connecting to outcomes
context, actions and outcomes
data collection and labeling
feedback loops
machine learning systems
properties
real usage
scale
simple interactions
TeamMaker
traditional computer vision system
unbiased data
understanding outcomes
escalations
implicit outcomes
reports
user classifications
user ratings
Data models
classification error
generalization error
regression error
Data pitfalls
bias
indirect value
privacy
rare events
Data, working with
conceptual machine learning
data models
pitfalls
broken confidence intervals
data bias
inconclusive data
noise in data
out of date
questions
structured data
Distribution of mistakes
E
E-commerce store
Evaluate intelligences
curves
generalization
mistakes
offline
online
operating point
predictions
probability
rankings
regressions
properties
subjective evaluations
Execution
change in right answer
conversion steps
intelligent experience occurring
Experience
levels, criteria
resources
F
False negative
False positive
Feature selection
Forceful automation
Forcefulness
Frequency
G
Generalization
Getting data
bootstrap data
computer vision case
learning curve
methods
usage data
ideal data-creation
options
H
Horror-game system
Human factor, intelligent mistakes
Hybrid experiences
I, J, K
Instability in intelligence
Intelligence
changes in
contexts
degradation
errors
management
pellet griller, example
predictions
classifications
hybrids and combinations
probability estimation
rankings
regressions
Intelligence creation
good creators
data debugging
intelligence-creation tools
math-first
verification-based approach
levels, criteria
process
achieving accurate objectives
blink detector example
efficiency, reliability
implementation, working
meaningful intelligence
phases
SeePhases, intelligence creation
tools
Intelligence experience
designing
position consideration
SeePosition consideration
positioning patterns
SeePatterns
Intelligence implementation
automatic updates
components
environment
management
orchestration
runtime
simple/complex
creators
customers
plug-in
user interaction
user-reported-offensive sites
web browsers
web page
web server
Intelligence orchestration
objectives
well-orchestrated intelligence
SeeOrchestration
Intelligence representation
SeeRepresentation
Intelligence system goals
achievable
anti-phishing
desired outcome
layering
measurement
A/B testing
ask the user
decoupling goals
hand labeling
processing
types
leading indicators
model properties
organizational objectives
user outcomes
Intelligence telemetry pipeline
Intelligent experiences
data collection
data creation, growth
effectiveness
failure reasons
flaws
intended
SeeIntended experience
mistakes
SeeMistakes
objectives
achieving system objectives
data creation
flaws
user presentation
options
system’s objectives
SeeIntelligent system, objectives
users
SeeUsers
verification, tools
Intelligent interaction modes
annotation
automated experience
hybrid experiences
organization
prompting action
Intelligent systems
building
checklist
SeeChecklist, intelligent system project
creation
data to toast
deciding, telemetry
elements
creation
experience
implementation
objective
orchestration
heuristic
Internet toaster
monitor, key metrics
objectives
design
home lighting automation
ineffective experience
leading indicators
minimize power amount
model properties
organizational outcomes
sensors
user outcomes
sensors
toast, machine learning
Intended experience
errors
user
working with context
computer vision system
crafting
Intelligent System state
producing manual contexts
recorded usage
work-flow
working with intelligence
Interactions levels, criteria
Intrinsically hard problems
Iris detector
L
Learning curve
Lighting up intelligence
controlled rollout
flighting
silent intelligence
single deployment
M
Machine learning
algorithms
complexity
complexity parameters
data size
feature
creation
engineering
selection
values
genre
hidden information
if-test
irrelevant features elimination
label
modeling
model search complexity
model structure complexity
normalization
overfitting
training set
underfits
Management
complexity
deploying the intelligence
frequency
human systems
lighting up intelligence
sanity checking
turning off intelligence
Mistakes
accuracy
artificial intelligence
changes
crazy mistakes
decisions
degradation
effectiveness
human factor
inaccurate intelligences
inconvenient
intelligence errors
interactions
machine learning
mitigation
balancing the experience
guardrails implementation
intelligence management parameters
investing in intelligence
override errors
model outages
perfection
pre-conceived notions
system outages
types
verification
Model outages
Model types
decision tree
forests
tree structure
linear model, working
neural networks
components
tasks
Monitoring success criteria
goals
levels
N
Normalization
O
One-hot encoding
Open-ended problems
Operating point
Operation
distribution costs, bandwidth charges
execution costs
offline
Orchestration
abuse
costs change
mistake costs
telemetry costs
importance
intelligence changes
objectives
opportunities
solving previous problems
understanding problems
problems
changing usage patterns
solving problems
time-changing problems
properties
team skills
types
balancing experience
creating intelligence
experience
inspecting interactions
intelligence creation
intelligence overriding
interactions
metrics dashboard
monitoring success criteria
telemetry sources
user changes
changing perception
changing usage patterns
new users
Organizing intelligence
comparing approaches
monolithic intelligences
properties
accuracy
comprehensible
growth
loose coupling
measurable
spaghetti code
team cohesion
reasons
cleaning mistakes
collaboration
incorporating legacy
problem solving
technique
SeeTechniques
Overriding intelligence
levels, criteria
work-flow management
P, Q
Passive experience
Patterns
back-end (cached) intelligence
advantages
disadvantages
client-side intelligence
disadvantages
extraction code
reasons, challenges
hybrid intelligence
server-centric intelligence, advantages
static intelligence
advantages
disadvantages
Phases, intelligence creation
data collection
methods
SeeGetting data
defining success
environments
low light sensor
questions
evaluation steps
heuristic creation
computer vision, techniques
methods
iteration
machine learning
complex artificial neural networks
machine-learning toolkit
standard approach
maturity creation
maturity spectrum
sanity-check
stages
tradeoffs, flexibility
Position consideration
latency
executing
updating
operation
SeeOperation
Precision-recall curve (PR curve)
Predictions
probability
rankings
regressions
Probability
Problem changing
Properties, intelligence
client-side summarization
flexible targeting
sampling
server-side summarization
R
Rankings
Receiver operating characteristic curve (ROC curve)
Regressions
Representation
code
heuristic-based intelligence
human-based intelligence
runtime performance
working
criteria, human, computer creation
data files
hand-labeling specific contexts
lookup tables, uses
machine learning
models
encode intelligence
types
SeeModel types
Runtime
context
effective
execution
feature extraction
information gathering
instability, intelligence
intelligence action
intelligence APIs
models
S
Sanity checking
compatibility
mistakes
runtime constraints
Scales, abuse
Internet
risk
communication
dedicated mind, hackers
mistakes
user generated content
users
Sub-populations evaluation
System outages
T
TeamMaker
Techniques
decoupled feature engineering
approach
challenges
meta-model
approaches
uses
mistakes
approaches
software bugs
model sequencing
approaches
uses
multiple model searches
approach
effectiveness
overriding
approaches
guardrails
hand-labeling specific contexts
uses
partition contexts
advantages
approaches
working
subjective scale
Telemetry system
data collection, intelligence
data pitfalls
functions
outcomes
properties
Testing set
Time-changing problems
Toaster-controlling algorithm
Toasters rule or update
Types of mistakes
U
Users
annotation
attention
automation
challenges
functioning
light-automation
light sensor reading
motion sensor data
organize
predictions
prompt
smart-home product
system’s goals
Uses, intelligent system
cost effective approaches
data usage
elements
partial system
system interface
types
big problems
intrinsically hard problems
open-ended problems
time-changing problems
V, W, X, Y, Z
Verification, intelligent experiences
continual
effectiveness
high-quality
Internet
measures
mistakes
poor-quality
scam sites
users
warning
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.